初始化项目文件

.dockerignore

@@ -0,0 +1,222 @@
+# Repo-specific DockerIgnore -------------------------------------------------------------------------------------------
+#.git
+.cache
+.idea
+runs
+output
+coco
+storage.googleapis.com
+
+data/samples/*
+**/results*.csv
+*.jpg
+
+# Neural Network weights -----------------------------------------------------------------------------------------------
+**/*.pt
+**/*.pth
+**/*.onnx
+**/*.engine
+**/*.mlmodel
+**/*.torchscript
+**/*.torchscript.pt
+**/*.tflite
+**/*.h5
+**/*.pb
+*_saved_model/
+*_web_model/
+*_openvino_model/
+
+# Below Copied From .gitignore -----------------------------------------------------------------------------------------
+# Below Copied From .gitignore -----------------------------------------------------------------------------------------
+
+
+# GitHub Python GitIgnore ----------------------------------------------------------------------------------------------
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+env/
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+*.egg-info/
+wandb/
+.installed.cfg
+*.egg
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+.hypothesis/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# pyenv
+.python-version
+
+# celery beat schedule file
+celerybeat-schedule
+
+# SageMath parsed files
+*.sage.py
+
+# dotenv
+.env
+
+# virtualenv
+.venv*
+venv*/
+ENV*/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+
+
+# https://github.com/github/gitignore/blob/master/Global/macOS.gitignore -----------------------------------------------
+
+# General
+.DS_Store
+.AppleDouble
+.LSOverride
+
+# Icon must end with two \r
+Icon
+Icon?
+
+# Thumbnails
+._*
+
+# Files that might appear in the root of a volume
+.DocumentRevisions-V100
+.fseventsd
+.Spotlight-V100
+.TemporaryItems
+.Trashes
+.VolumeIcon.icns
+.com.apple.timemachine.donotpresent
+
+# Directories potentially created on remote AFP share
+.AppleDB
+.AppleDesktop
+Network Trash Folder
+Temporary Items
+.apdisk
+
+
+# https://github.com/github/gitignore/blob/master/Global/JetBrains.gitignore
+# Covers JetBrains IDEs: IntelliJ, RubyMine, PhpStorm, AppCode, PyCharm, CLion, Android Studio and WebStorm
+# Reference: https://intellij-support.jetbrains.com/hc/en-us/articles/206544839
+
+# User-specific stuff:
+.idea/*
+.idea/**/workspace.xml
+.idea/**/tasks.xml
+.idea/dictionaries
+.html  # Bokeh Plots
+.pg  # TensorFlow Frozen Graphs
+.avi # videos
+
+# Sensitive or high-churn files:
+.idea/**/dataSources/
+.idea/**/dataSources.ids
+.idea/**/dataSources.local.xml
+.idea/**/sqlDataSources.xml
+.idea/**/dynamic.xml
+.idea/**/uiDesigner.xml
+
+# Gradle:
+.idea/**/gradle.xml
+.idea/**/libraries
+
+# CMake
+cmake-build-debug/
+cmake-build-release/
+
+# Mongo Explorer plugin:
+.idea/**/mongoSettings.xml
+
+## File-based project format:
+*.iws
+
+## Plugin-specific files:
+
+# IntelliJ
+out/
+
+# mpeltonen/sbt-idea plugin
+.idea_modules/
+
+# JIRA plugin
+atlassian-ide-plugin.xml
+
+# Cursive Clojure plugin
+.idea/replstate.xml
+
+# Crashlytics plugin (for Android Studio and IntelliJ)
+com_crashlytics_export_strings.xml
+crashlytics.properties
+crashlytics-build.properties
+fabric.properties

.gitattributes

@@ -0,0 +1,2 @@
+# this drop notebooks from GitHub language stats
+*.ipynb linguist-vendored

.gitignore

@@ -41,6 +41,9 @@ results*.csv
 coco/
 coco128/
 VOC/
+VOC2007_wm
+datasets
+.github
 
 # MATLAB GitIgnore -----------------------------------------------------------------------------------------------------
 *.m~

.pre-commit-config.yaml

@@ -0,0 +1,66 @@
+# Define hooks for code formations
+# Will be applied on any updated commit files if a user has installed and linked commit hook
+
+default_language_version:
+  python: python3.8
+
+# Define bot property if installed via https://github.com/marketplace/pre-commit-ci
+ci:
+  autofix_prs: true
+  autoupdate_commit_msg: '[pre-commit.ci] pre-commit suggestions'
+  autoupdate_schedule: quarterly
+  # submodules: true
+
+repos:
+  - repo: https://github.com/pre-commit/pre-commit-hooks
+    rev: v4.1.0
+    hooks:
+      - id: end-of-file-fixer
+      - id: trailing-whitespace
+      - id: check-case-conflict
+      - id: check-yaml
+      - id: check-toml
+      - id: pretty-format-json
+      - id: check-docstring-first
+
+  - repo: https://github.com/asottile/pyupgrade
+    rev: v2.31.0
+    hooks:
+      - id: pyupgrade
+        args: [--py36-plus]
+        name: Upgrade code
+
+  - repo: https://github.com/PyCQA/isort
+    rev: 5.10.1
+    hooks:
+      - id: isort
+        name: Sort imports
+
+  # TODO
+  #- repo: https://github.com/pre-commit/mirrors-yapf
+  #  rev: v0.31.0
+  #  hooks:
+  #    - id: yapf
+  #      name: formatting
+
+  # TODO
+  #- repo: https://github.com/executablebooks/mdformat
+  #  rev: 0.7.7
+  #  hooks:
+  #    - id: mdformat
+  #      additional_dependencies:
+  #        - mdformat-gfm
+  #        - mdformat-black
+  #        - mdformat_frontmatter
+
+  # TODO
+  #- repo: https://github.com/asottile/yesqa
+  #  rev: v1.2.3
+  #  hooks:
+  #    - id: yesqa
+
+  - repo: https://github.com/PyCQA/flake8
+    rev: 4.0.1
+    hooks:
+      - id: flake8
+        name: PEP8

CONTRIBUTING.md

@@ -0,0 +1,94 @@
+## Contributing to YOLOv5 🚀
+
+We love your input! We want to make contributing to YOLOv5 as easy and transparent as possible, whether it's:
+
+- Reporting a bug
+- Discussing the current state of the code
+- Submitting a fix
+- Proposing a new feature
+- Becoming a maintainer
+
+YOLOv5 works so well due to our combined community effort, and for every small improvement you contribute you will be
+helping push the frontiers of what's possible in AI 😃!
+
+## Submitting a Pull Request (PR) 🛠️
+
+Submitting a PR is easy! This example shows how to submit a PR for updating `requirements.txt` in 4 steps:
+
+### 1. Select File to Update
+
+Select `requirements.txt` to update by clicking on it in GitHub.
+<p align="center"><img width="800" alt="PR_step1" src="https://user-images.githubusercontent.com/26833433/122260847-08be2600-ced4-11eb-828b-8287ace4136c.png"></p>
+
+### 2. Click 'Edit this file'
+
+Button is in top-right corner.
+<p align="center"><img width="800" alt="PR_step2" src="https://user-images.githubusercontent.com/26833433/122260844-06f46280-ced4-11eb-9eec-b8a24be519ca.png"></p>
+
+### 3. Make Changes
+
+Change `matplotlib` version from `3.2.2` to `3.3`.
+<p align="center"><img width="800" alt="PR_step3" src="https://user-images.githubusercontent.com/26833433/122260853-0a87e980-ced4-11eb-9fd2-3650fb6e0842.png"></p>
+
+### 4. Preview Changes and Submit PR
+
+Click on the **Preview changes** tab to verify your updates. At the bottom of the screen select 'Create a **new branch**
+for this commit', assign your branch a descriptive name such as `fix/matplotlib_version` and click the green **Propose
+changes** button. All done, your PR is now submitted to YOLOv5 for review and approval 😃!
+<p align="center"><img width="800" alt="PR_step4" src="https://user-images.githubusercontent.com/26833433/122260856-0b208000-ced4-11eb-8e8e-77b6151cbcc3.png"></p>
+
+### PR recommendations
+
+To allow your work to be integrated as seamlessly as possible, we advise you to:
+
+- ✅ Verify your PR is **up-to-date with upstream/master.** If your PR is behind upstream/master an
+  automatic [GitHub Actions](https://github.com/ultralytics/yolov5/blob/master/.github/workflows/rebase.yml) merge may
+  be attempted by writing /rebase in a new comment, or by running the following code, replacing 'feature' with the name
+  of your local branch:
+
+```bash
+git remote add upstream https://github.com/ultralytics/yolov5.git
+git fetch upstream
+# git checkout feature  # <--- replace 'feature' with local branch name
+git merge upstream/master
+git push -u origin -f
+```
+
+- ✅ Verify all Continuous Integration (CI) **checks are passing**.
+- ✅ Reduce changes to the absolute **minimum** required for your bug fix or feature addition. _"It is not daily increase
+  but daily decrease, hack away the unessential. The closer to the source, the less wastage there is."_  — Bruce Lee
+
+## Submitting a Bug Report 🐛
+
+If you spot a problem with YOLOv5 please submit a Bug Report!
+
+For us to start investigating a possible problem we need to be able to reproduce it ourselves first. We've created a few
+short guidelines below to help users provide what we need in order to get started.
+
+When asking a question, people will be better able to provide help if you provide **code** that they can easily
+understand and use to **reproduce** the problem. This is referred to by community members as creating
+a [minimum reproducible example](https://stackoverflow.com/help/minimal-reproducible-example). Your code that reproduces
+the problem should be:
+
+* ✅ **Minimal** – Use as little code as possible that still produces the same problem
+* ✅ **Complete** – Provide **all** parts someone else needs to reproduce your problem in the question itself
+* ✅ **Reproducible** – Test the code you're about to provide to make sure it reproduces the problem
+
+In addition to the above requirements, for [Ultralytics](https://ultralytics.com/) to provide assistance your code
+should be:
+
+* ✅ **Current** – Verify that your code is up-to-date with current
+  GitHub [master](https://github.com/ultralytics/yolov5/tree/master), and if necessary `git pull` or `git clone` a new
+  copy to ensure your problem has not already been resolved by previous commits.
+* ✅ **Unmodified** – Your problem must be reproducible without any modifications to the codebase in this
+  repository. [Ultralytics](https://ultralytics.com/) does not provide support for custom code ⚠️.
+
+If you believe your problem meets all of the above criteria, please close this issue and raise a new one using the 🐛 **
+Bug Report** [template](https://github.com/ultralytics/yolov5/issues/new/choose) and providing
+a [minimum reproducible example](https://stackoverflow.com/help/minimal-reproducible-example) to help us better
+understand and diagnose your problem.
+
+## License
+
+By contributing, you agree that your contributions will be licensed under
+the [GPL-3.0 license](https://choosealicense.com/licenses/gpl-3.0/)

Dockerfile

@@ -0,0 +1,64 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+
+# Start FROM Nvidia PyTorch image https://ngc.nvidia.com/catalog/containers/nvidia:pytorch
+FROM nvcr.io/nvidia/pytorch:21.10-py3
+
+# Install linux packages
+RUN apt update && apt install -y zip htop screen libgl1-mesa-glx
+
+# Install python dependencies
+COPY requirements.txt .
+RUN python -m pip install --upgrade pip
+RUN pip uninstall -y torch torchvision torchtext
+RUN pip install --no-cache -r requirements.txt albumentations wandb gsutil notebook \
+    torch==1.10.2+cu113 torchvision==0.11.3+cu113 -f https://download.pytorch.org/whl/cu113/torch_stable.html
+# RUN pip install --no-cache -U torch torchvision
+
+# Create working directory
+RUN mkdir -p /usr/src/app
+WORKDIR /usr/src/app
+
+# Copy contents
+COPY . /usr/src/app
+
+# Downloads to user config dir
+ADD https://ultralytics.com/assets/Arial.ttf /root/.config/Ultralytics/
+
+# Set environment variables
+# ENV HOME=/usr/src/app
+
+
+# Usage Examples -------------------------------------------------------------------------------------------------------
+
+# Build and Push
+# t=ultralytics/yolov5:latest && sudo docker build -t $t . && sudo docker push $t
+
+# Pull and Run
+# t=ultralytics/yolov5:latest && sudo docker pull $t && sudo docker run -it --ipc=host --gpus all $t
+
+# Pull and Run with local directory access
+# t=ultralytics/yolov5:latest && sudo docker pull $t && sudo docker run -it --ipc=host --gpus all -v "$(pwd)"/datasets:/usr/src/datasets $t
+
+# Kill all
+# sudo docker kill $(sudo docker ps -q)
+
+# Kill all image-based
+# sudo docker kill $(sudo docker ps -qa --filter ancestor=ultralytics/yolov5:latest)
+
+# Bash into running container
+# sudo docker exec -it 5a9b5863d93d bash
+
+# Bash into stopped container
+# id=$(sudo docker ps -qa) && sudo docker start $id && sudo docker exec -it $id bash
+
+# Clean up
+# docker system prune -a --volumes
+
+# Update Ubuntu drivers
+# https://www.maketecheasier.com/install-nvidia-drivers-ubuntu/
+
+# DDP test
+# python -m torch.distributed.run --nproc_per_node 2 --master_port 1 train.py --epochs 3
+
+# GCP VM from Image
+# docker.io/ultralytics/yolov5:latest

LICENSE

@@ -0,0 +1,674 @@
+GNU GENERAL PUBLIC LICENSE
+                       Version 3, 29 June 2007
+
+ Copyright (C) 2007 Free Software Foundation, Inc. <http://fsf.org/>
+ Everyone is permitted to copy and distribute verbatim copies
+ of this license document, but changing it is not allowed.
+
+                            Preamble
+
+  The GNU General Public License is a free, copyleft license for
+software and other kinds of works.
+
+  The licenses for most software and other practical works are designed
+to take away your freedom to share and change the works.  By contrast,
+the GNU General Public License is intended to guarantee your freedom to
+share and change all versions of a program--to make sure it remains free
+software for all its users.  We, the Free Software Foundation, use the
+GNU General Public License for most of our software; it applies also to
+any other work released this way by its authors.  You can apply it to
+your programs, too.
+
+  When we speak of free software, we are referring to freedom, not
+price.  Our General Public Licenses are designed to make sure that you
+have the freedom to distribute copies of free software (and charge for
+them if you wish), that you receive source code or can get it if you
+want it, that you can change the software or use pieces of it in new
+free programs, and that you know you can do these things.
+
+  To protect your rights, we need to prevent others from denying you
+these rights or asking you to surrender the rights.  Therefore, you have
+certain responsibilities if you distribute copies of the software, or if
+you modify it: responsibilities to respect the freedom of others.
+
+  For example, if you distribute copies of such a program, whether
+gratis or for a fee, you must pass on to the recipients the same
+freedoms that you received.  You must make sure that they, too, receive
+or can get the source code.  And you must show them these terms so they
+know their rights.
+
+  Developers that use the GNU GPL protect your rights with two steps:
+(1) assert copyright on the software, and (2) offer you this License
+giving you legal permission to copy, distribute and/or modify it.
+
+  For the developers' and authors' protection, the GPL clearly explains
+that there is no warranty for this free software.  For both users' and
+authors' sake, the GPL requires that modified versions be marked as
+changed, so that their problems will not be attributed erroneously to
+authors of previous versions.
+
+  Some devices are designed to deny users access to install or run
+modified versions of the software inside them, although the manufacturer
+can do so.  This is fundamentally incompatible with the aim of
+protecting users' freedom to change the software.  The systematic
+pattern of such abuse occurs in the area of products for individuals to
+use, which is precisely where it is most unacceptable.  Therefore, we
+have designed this version of the GPL to prohibit the practice for those
+products.  If such problems arise substantially in other domains, we
+stand ready to extend this provision to those domains in future versions
+of the GPL, as needed to protect the freedom of users.
+
+  Finally, every program is threatened constantly by software patents.
+States should not allow patents to restrict development and use of
+software on general-purpose computers, but in those that do, we wish to
+avoid the special danger that patents applied to a free program could
+make it effectively proprietary.  To prevent this, the GPL assures that
+patents cannot be used to render the program non-free.
+
+  The precise terms and conditions for copying, distribution and
+modification follow.
+
+                       TERMS AND CONDITIONS
+
+  0. Definitions.
+
+  "This License" refers to version 3 of the GNU General Public License.
+
+  "Copyright" also means copyright-like laws that apply to other kinds of
+works, such as semiconductor masks.
+
+  "The Program" refers to any copyrightable work licensed under this
+License.  Each licensee is addressed as "you".  "Licensees" and
+"recipients" may be individuals or organizations.
+
+  To "modify" a work means to copy from or adapt all or part of the work
+in a fashion requiring copyright permission, other than the making of an
+exact copy.  The resulting work is called a "modified version" of the
+earlier work or a work "based on" the earlier work.
+
+  A "covered work" means either the unmodified Program or a work based
+on the Program.
+
+  To "propagate" a work means to do anything with it that, without
+permission, would make you directly or secondarily liable for
+infringement under applicable copyright law, except executing it on a
+computer or modifying a private copy.  Propagation includes copying,
+distribution (with or without modification), making available to the
+public, and in some countries other activities as well.
+
+  To "convey" a work means any kind of propagation that enables other
+parties to make or receive copies.  Mere interaction with a user through
+a computer network, with no transfer of a copy, is not conveying.
+
+  An interactive user interface displays "Appropriate Legal Notices"
+to the extent that it includes a convenient and prominently visible
+feature that (1) displays an appropriate copyright notice, and (2)
+tells the user that there is no warranty for the work (except to the
+extent that warranties are provided), that licensees may convey the
+work under this License, and how to view a copy of this License.  If
+the interface presents a list of user commands or options, such as a
+menu, a prominent item in the list meets this criterion.
+
+  1. Source Code.
+
+  The "source code" for a work means the preferred form of the work
+for making modifications to it.  "Object code" means any non-source
+form of a work.
+
+  A "Standard Interface" means an interface that either is an official
+standard defined by a recognized standards body, or, in the case of
+interfaces specified for a particular programming language, one that
+is widely used among developers working in that language.
+
+  The "System Libraries" of an executable work include anything, other
+than the work as a whole, that (a) is included in the normal form of
+packaging a Major Component, but which is not part of that Major
+Component, and (b) serves only to enable use of the work with that
+Major Component, or to implement a Standard Interface for which an
+implementation is available to the public in source code form.  A
+"Major Component", in this context, means a major essential component
+(kernel, window system, and so on) of the specific operating system
+(if any) on which the executable work runs, or a compiler used to
+produce the work, or an object code interpreter used to run it.
+
+  The "Corresponding Source" for a work in object code form means all
+the source code needed to generate, install, and (for an executable
+work) run the object code and to modify the work, including scripts to
+control those activities.  However, it does not include the work's
+System Libraries, or general-purpose tools or generally available free
+programs which are used unmodified in performing those activities but
+which are not part of the work.  For example, Corresponding Source
+includes interface definition files associated with source files for
+the work, and the source code for shared libraries and dynamically
+linked subprograms that the work is specifically designed to require,
+such as by intimate data communication or control flow between those
+subprograms and other parts of the work.
+
+  The Corresponding Source need not include anything that users
+can regenerate automatically from other parts of the Corresponding
+Source.
+
+  The Corresponding Source for a work in source code form is that
+same work.
+
+  2. Basic Permissions.
+
+  All rights granted under this License are granted for the term of
+copyright on the Program, and are irrevocable provided the stated
+conditions are met.  This License explicitly affirms your unlimited
+permission to run the unmodified Program.  The output from running a
+covered work is covered by this License only if the output, given its
+content, constitutes a covered work.  This License acknowledges your
+rights of fair use or other equivalent, as provided by copyright law.
+
+  You may make, run and propagate covered works that you do not
+convey, without conditions so long as your license otherwise remains
+in force.  You may convey covered works to others for the sole purpose
+of having them make modifications exclusively for you, or provide you
+with facilities for running those works, provided that you comply with
+the terms of this License in conveying all material for which you do
+not control copyright.  Those thus making or running the covered works
+for you must do so exclusively on your behalf, under your direction
+and control, on terms that prohibit them from making any copies of
+your copyrighted material outside their relationship with you.
+
+  Conveying under any other circumstances is permitted solely under
+the conditions stated below.  Sublicensing is not allowed; section 10
+makes it unnecessary.
+
+  3. Protecting Users' Legal Rights From Anti-Circumvention Law.
+
+  No covered work shall be deemed part of an effective technological
+measure under any applicable law fulfilling obligations under article
+11 of the WIPO copyright treaty adopted on 20 December 1996, or
+similar laws prohibiting or restricting circumvention of such
+measures.
+
+  When you convey a covered work, you waive any legal power to forbid
+circumvention of technological measures to the extent such circumvention
+is effected by exercising rights under this License with respect to
+the covered work, and you disclaim any intention to limit operation or
+modification of the work as a means of enforcing, against the work's
+users, your or third parties' legal rights to forbid circumvention of
+technological measures.
+
+  4. Conveying Verbatim Copies.
+
+  You may convey verbatim copies of the Program's source code as you
+receive it, in any medium, provided that you conspicuously and
+appropriately publish on each copy an appropriate copyright notice;
+keep intact all notices stating that this License and any
+non-permissive terms added in accord with section 7 apply to the code;
+keep intact all notices of the absence of any warranty; and give all
+recipients a copy of this License along with the Program.
+
+  You may charge any price or no price for each copy that you convey,
+and you may offer support or warranty protection for a fee.
+
+  5. Conveying Modified Source Versions.
+
+  You may convey a work based on the Program, or the modifications to
+produce it from the Program, in the form of source code under the
+terms of section 4, provided that you also meet all of these conditions:
+
+    a) The work must carry prominent notices stating that you modified
+    it, and giving a relevant date.
+
+    b) The work must carry prominent notices stating that it is
+    released under this License and any conditions added under section
+    7.  This requirement modifies the requirement in section 4 to
+    "keep intact all notices".
+
+    c) You must license the entire work, as a whole, under this
+    License to anyone who comes into possession of a copy.  This
+    License will therefore apply, along with any applicable section 7
+    additional terms, to the whole of the work, and all its parts,
+    regardless of how they are packaged.  This License gives no
+    permission to license the work in any other way, but it does not
+    invalidate such permission if you have separately received it.
+
+    d) If the work has interactive user interfaces, each must display
+    Appropriate Legal Notices; however, if the Program has interactive
+    interfaces that do not display Appropriate Legal Notices, your
+    work need not make them do so.
+
+  A compilation of a covered work with other separate and independent
+works, which are not by their nature extensions of the covered work,
+and which are not combined with it such as to form a larger program,
+in or on a volume of a storage or distribution medium, is called an
+"aggregate" if the compilation and its resulting copyright are not
+used to limit the access or legal rights of the compilation's users
+beyond what the individual works permit.  Inclusion of a covered work
+in an aggregate does not cause this License to apply to the other
+parts of the aggregate.
+
+  6. Conveying Non-Source Forms.
+
+  You may convey a covered work in object code form under the terms
+of sections 4 and 5, provided that you also convey the
+machine-readable Corresponding Source under the terms of this License,
+in one of these ways:
+
+    a) Convey the object code in, or embodied in, a physical product
+    (including a physical distribution medium), accompanied by the
+    Corresponding Source fixed on a durable physical medium
+    customarily used for software interchange.
+
+    b) Convey the object code in, or embodied in, a physical product
+    (including a physical distribution medium), accompanied by a
+    written offer, valid for at least three years and valid for as
+    long as you offer spare parts or customer support for that product
+    model, to give anyone who possesses the object code either (1) a
+    copy of the Corresponding Source for all the software in the
+    product that is covered by this License, on a durable physical
+    medium customarily used for software interchange, for a price no
+    more than your reasonable cost of physically performing this
+    conveying of source, or (2) access to copy the
+    Corresponding Source from a network server at no charge.
+
+    c) Convey individual copies of the object code with a copy of the
+    written offer to provide the Corresponding Source.  This
+    alternative is allowed only occasionally and noncommercially, and
+    only if you received the object code with such an offer, in accord
+    with subsection 6b.
+
+    d) Convey the object code by offering access from a designated
+    place (gratis or for a charge), and offer equivalent access to the
+    Corresponding Source in the same way through the same place at no
+    further charge.  You need not require recipients to copy the
+    Corresponding Source along with the object code.  If the place to
+    copy the object code is a network server, the Corresponding Source
+    may be on a different server (operated by you or a third party)
+    that supports equivalent copying facilities, provided you maintain
+    clear directions next to the object code saying where to find the
+    Corresponding Source.  Regardless of what server hosts the
+    Corresponding Source, you remain obligated to ensure that it is
+    available for as long as needed to satisfy these requirements.
+
+    e) Convey the object code using peer-to-peer transmission, provided
+    you inform other peers where the object code and Corresponding
+    Source of the work are being offered to the general public at no
+    charge under subsection 6d.
+
+  A separable portion of the object code, whose source code is excluded
+from the Corresponding Source as a System Library, need not be
+included in conveying the object code work.
+
+  A "User Product" is either (1) a "consumer product", which means any
+tangible personal property which is normally used for personal, family,
+or household purposes, or (2) anything designed or sold for incorporation
+into a dwelling.  In determining whether a product is a consumer product,
+doubtful cases shall be resolved in favor of coverage.  For a particular
+product received by a particular user, "normally used" refers to a
+typical or common use of that class of product, regardless of the status
+of the particular user or of the way in which the particular user
+actually uses, or expects or is expected to use, the product.  A product
+is a consumer product regardless of whether the product has substantial
+commercial, industrial or non-consumer uses, unless such uses represent
+the only significant mode of use of the product.
+
+  "Installation Information" for a User Product means any methods,
+procedures, authorization keys, or other information required to install
+and execute modified versions of a covered work in that User Product from
+a modified version of its Corresponding Source.  The information must
+suffice to ensure that the continued functioning of the modified object
+code is in no case prevented or interfered with solely because
+modification has been made.
+
+  If you convey an object code work under this section in, or with, or
+specifically for use in, a User Product, and the conveying occurs as
+part of a transaction in which the right of possession and use of the
+User Product is transferred to the recipient in perpetuity or for a
+fixed term (regardless of how the transaction is characterized), the
+Corresponding Source conveyed under this section must be accompanied
+by the Installation Information.  But this requirement does not apply
+if neither you nor any third party retains the ability to install
+modified object code on the User Product (for example, the work has
+been installed in ROM).
+
+  The requirement to provide Installation Information does not include a
+requirement to continue to provide support service, warranty, or updates
+for a work that has been modified or installed by the recipient, or for
+the User Product in which it has been modified or installed.  Access to a
+network may be denied when the modification itself materially and
+adversely affects the operation of the network or violates the rules and
+protocols for communication across the network.
+
+  Corresponding Source conveyed, and Installation Information provided,
+in accord with this section must be in a format that is publicly
+documented (and with an implementation available to the public in
+source code form), and must require no special password or key for
+unpacking, reading or copying.
+
+  7. Additional Terms.
+
+  "Additional permissions" are terms that supplement the terms of this
+License by making exceptions from one or more of its conditions.
+Additional permissions that are applicable to the entire Program shall
+be treated as though they were included in this License, to the extent
+that they are valid under applicable law.  If additional permissions
+apply only to part of the Program, that part may be used separately
+under those permissions, but the entire Program remains governed by
+this License without regard to the additional permissions.
+
+  When you convey a copy of a covered work, you may at your option
+remove any additional permissions from that copy, or from any part of
+it.  (Additional permissions may be written to require their own
+removal in certain cases when you modify the work.)  You may place
+additional permissions on material, added by you to a covered work,
+for which you have or can give appropriate copyright permission.
+
+  Notwithstanding any other provision of this License, for material you
+add to a covered work, you may (if authorized by the copyright holders of
+that material) supplement the terms of this License with terms:
+
+    a) Disclaiming warranty or limiting liability differently from the
+    terms of sections 15 and 16 of this License; or
+
+    b) Requiring preservation of specified reasonable legal notices or
+    author attributions in that material or in the Appropriate Legal
+    Notices displayed by works containing it; or
+
+    c) Prohibiting misrepresentation of the origin of that material, or
+    requiring that modified versions of such material be marked in
+    reasonable ways as different from the original version; or
+
+    d) Limiting the use for publicity purposes of names of licensors or
+    authors of the material; or
+
+    e) Declining to grant rights under trademark law for use of some
+    trade names, trademarks, or service marks; or
+
+    f) Requiring indemnification of licensors and authors of that
+    material by anyone who conveys the material (or modified versions of
+    it) with contractual assumptions of liability to the recipient, for
+    any liability that these contractual assumptions directly impose on
+    those licensors and authors.
+
+  All other non-permissive additional terms are considered "further
+restrictions" within the meaning of section 10.  If the Program as you
+received it, or any part of it, contains a notice stating that it is
+governed by this License along with a term that is a further
+restriction, you may remove that term.  If a license document contains
+a further restriction but permits relicensing or conveying under this
+License, you may add to a covered work material governed by the terms
+of that license document, provided that the further restriction does
+not survive such relicensing or conveying.
+
+  If you add terms to a covered work in accord with this section, you
+must place, in the relevant source files, a statement of the
+additional terms that apply to those files, or a notice indicating
+where to find the applicable terms.
+
+  Additional terms, permissive or non-permissive, may be stated in the
+form of a separately written license, or stated as exceptions;
+the above requirements apply either way.
+
+  8. Termination.
+
+  You may not propagate or modify a covered work except as expressly
+provided under this License.  Any attempt otherwise to propagate or
+modify it is void, and will automatically terminate your rights under
+this License (including any patent licenses granted under the third
+paragraph of section 11).
+
+  However, if you cease all violation of this License, then your
+license from a particular copyright holder is reinstated (a)
+provisionally, unless and until the copyright holder explicitly and
+finally terminates your license, and (b) permanently, if the copyright
+holder fails to notify you of the violation by some reasonable means
+prior to 60 days after the cessation.
+
+  Moreover, your license from a particular copyright holder is
+reinstated permanently if the copyright holder notifies you of the
+violation by some reasonable means, this is the first time you have
+received notice of violation of this License (for any work) from that
+copyright holder, and you cure the violation prior to 30 days after
+your receipt of the notice.
+
+  Termination of your rights under this section does not terminate the
+licenses of parties who have received copies or rights from you under
+this License.  If your rights have been terminated and not permanently
+reinstated, you do not qualify to receive new licenses for the same
+material under section 10.
+
+  9. Acceptance Not Required for Having Copies.
+
+  You are not required to accept this License in order to receive or
+run a copy of the Program.  Ancillary propagation of a covered work
+occurring solely as a consequence of using peer-to-peer transmission
+to receive a copy likewise does not require acceptance.  However,
+nothing other than this License grants you permission to propagate or
+modify any covered work.  These actions infringe copyright if you do
+not accept this License.  Therefore, by modifying or propagating a
+covered work, you indicate your acceptance of this License to do so.
+
+  10. Automatic Licensing of Downstream Recipients.
+
+  Each time you convey a covered work, the recipient automatically
+receives a license from the original licensors, to run, modify and
+propagate that work, subject to this License.  You are not responsible
+for enforcing compliance by third parties with this License.
+
+  An "entity transaction" is a transaction transferring control of an
+organization, or substantially all assets of one, or subdividing an
+organization, or merging organizations.  If propagation of a covered
+work results from an entity transaction, each party to that
+transaction who receives a copy of the work also receives whatever
+licenses to the work the party's predecessor in interest had or could
+give under the previous paragraph, plus a right to possession of the
+Corresponding Source of the work from the predecessor in interest, if
+the predecessor has it or can get it with reasonable efforts.
+
+  You may not impose any further restrictions on the exercise of the
+rights granted or affirmed under this License.  For example, you may
+not impose a license fee, royalty, or other charge for exercise of
+rights granted under this License, and you may not initiate litigation
+(including a cross-claim or counterclaim in a lawsuit) alleging that
+any patent claim is infringed by making, using, selling, offering for
+sale, or importing the Program or any portion of it.
+
+  11. Patents.
+
+  A "contributor" is a copyright holder who authorizes use under this
+License of the Program or a work on which the Program is based.  The
+work thus licensed is called the contributor's "contributor version".
+
+  A contributor's "essential patent claims" are all patent claims
+owned or controlled by the contributor, whether already acquired or
+hereafter acquired, that would be infringed by some manner, permitted
+by this License, of making, using, or selling its contributor version,
+but do not include claims that would be infringed only as a
+consequence of further modification of the contributor version.  For
+purposes of this definition, "control" includes the right to grant
+patent sublicenses in a manner consistent with the requirements of
+this License.
+
+  Each contributor grants you a non-exclusive, worldwide, royalty-free
+patent license under the contributor's essential patent claims, to
+make, use, sell, offer for sale, import and otherwise run, modify and
+propagate the contents of its contributor version.
+
+  In the following three paragraphs, a "patent license" is any express
+agreement or commitment, however denominated, not to enforce a patent
+(such as an express permission to practice a patent or covenant not to
+sue for patent infringement).  To "grant" such a patent license to a
+party means to make such an agreement or commitment not to enforce a
+patent against the party.
+
+  If you convey a covered work, knowingly relying on a patent license,
+and the Corresponding Source of the work is not available for anyone
+to copy, free of charge and under the terms of this License, through a
+publicly available network server or other readily accessible means,
+then you must either (1) cause the Corresponding Source to be so
+available, or (2) arrange to deprive yourself of the benefit of the
+patent license for this particular work, or (3) arrange, in a manner
+consistent with the requirements of this License, to extend the patent
+license to downstream recipients.  "Knowingly relying" means you have
+actual knowledge that, but for the patent license, your conveying the
+covered work in a country, or your recipient's use of the covered work
+in a country, would infringe one or more identifiable patents in that
+country that you have reason to believe are valid.
+
+  If, pursuant to or in connection with a single transaction or
+arrangement, you convey, or propagate by procuring conveyance of, a
+covered work, and grant a patent license to some of the parties
+receiving the covered work authorizing them to use, propagate, modify
+or convey a specific copy of the covered work, then the patent license
+you grant is automatically extended to all recipients of the covered
+work and works based on it.
+
+  A patent license is "discriminatory" if it does not include within
+the scope of its coverage, prohibits the exercise of, or is
+conditioned on the non-exercise of one or more of the rights that are
+specifically granted under this License.  You may not convey a covered
+work if you are a party to an arrangement with a third party that is
+in the business of distributing software, under which you make payment
+to the third party based on the extent of your activity of conveying
+the work, and under which the third party grants, to any of the
+parties who would receive the covered work from you, a discriminatory
+patent license (a) in connection with copies of the covered work
+conveyed by you (or copies made from those copies), or (b) primarily
+for and in connection with specific products or compilations that
+contain the covered work, unless you entered into that arrangement,
+or that patent license was granted, prior to 28 March 2007.
+
+  Nothing in this License shall be construed as excluding or limiting
+any implied license or other defenses to infringement that may
+otherwise be available to you under applicable patent law.
+
+  12. No Surrender of Others' Freedom.
+
+  If conditions are imposed on you (whether by court order, agreement or
+otherwise) that contradict the conditions of this License, they do not
+excuse you from the conditions of this License.  If you cannot convey a
+covered work so as to satisfy simultaneously your obligations under this
+License and any other pertinent obligations, then as a consequence you may
+not convey it at all.  For example, if you agree to terms that obligate you
+to collect a royalty for further conveying from those to whom you convey
+the Program, the only way you could satisfy both those terms and this
+License would be to refrain entirely from conveying the Program.
+
+  13. Use with the GNU Affero General Public License.
+
+  Notwithstanding any other provision of this License, you have
+permission to link or combine any covered work with a work licensed
+under version 3 of the GNU Affero General Public License into a single
+combined work, and to convey the resulting work.  The terms of this
+License will continue to apply to the part which is the covered work,
+but the special requirements of the GNU Affero General Public License,
+section 13, concerning interaction through a network will apply to the
+combination as such.
+
+  14. Revised Versions of this License.
+
+  The Free Software Foundation may publish revised and/or new versions of
+the GNU General Public License from time to time.  Such new versions will
+be similar in spirit to the present version, but may differ in detail to
+address new problems or concerns.
+
+  Each version is given a distinguishing version number.  If the
+Program specifies that a certain numbered version of the GNU General
+Public License "or any later version" applies to it, you have the
+option of following the terms and conditions either of that numbered
+version or of any later version published by the Free Software
+Foundation.  If the Program does not specify a version number of the
+GNU General Public License, you may choose any version ever published
+by the Free Software Foundation.
+
+  If the Program specifies that a proxy can decide which future
+versions of the GNU General Public License can be used, that proxy's
+public statement of acceptance of a version permanently authorizes you
+to choose that version for the Program.
+
+  Later license versions may give you additional or different
+permissions.  However, no additional obligations are imposed on any
+author or copyright holder as a result of your choosing to follow a
+later version.
+
+  15. Disclaimer of Warranty.
+
+  THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY
+APPLICABLE LAW.  EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT
+HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY
+OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO,
+THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+PURPOSE.  THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM
+IS WITH YOU.  SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF
+ALL NECESSARY SERVICING, REPAIR OR CORRECTION.
+
+  16. Limitation of Liability.
+
+  IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
+WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS
+THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY
+GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE
+USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF
+DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD
+PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS),
+EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF
+SUCH DAMAGES.
+
+  17. Interpretation of Sections 15 and 16.
+
+  If the disclaimer of warranty and limitation of liability provided
+above cannot be given local legal effect according to their terms,
+reviewing courts shall apply local law that most closely approximates
+an absolute waiver of all civil liability in connection with the
+Program, unless a warranty or assumption of liability accompanies a
+copy of the Program in return for a fee.
+
+                     END OF TERMS AND CONDITIONS
+
+            How to Apply These Terms to Your New Programs
+
+  If you develop a new program, and you want it to be of the greatest
+possible use to the public, the best way to achieve this is to make it
+free software which everyone can redistribute and change under these terms.
+
+  To do so, attach the following notices to the program.  It is safest
+to attach them to the start of each source file to most effectively
+state the exclusion of warranty; and each file should have at least
+the "copyright" line and a pointer to where the full notice is found.
+
+    <one line to give the program's name and a brief idea of what it does.>
+    Copyright (C) <year>  <name of author>
+
+    This program is free software: you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation, either version 3 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License
+    along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+Also add information on how to contact you by electronic and paper mail.
+
+  If the program does terminal interaction, make it output a short
+notice like this when it starts in an interactive mode:
+
+    <program>  Copyright (C) <year>  <name of author>
+    This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'.
+    This is free software, and you are welcome to redistribute it
+    under certain conditions; type `show c' for details.
+
+The hypothetical commands `show w' and `show c' should show the appropriate
+parts of the General Public License.  Of course, your program's commands
+might be different; for a GUI interface, you would use an "about box".
+
+  You should also get your employer (if you work as a programmer) or school,
+if any, to sign a "copyright disclaimer" for the program, if necessary.
+For more information on this, and how to apply and follow the GNU GPL, see
+<http://www.gnu.org/licenses/>.
+
+  The GNU General Public License does not permit incorporating your program
+into proprietary programs.  If your program is a subroutine library, you
+may consider it more useful to permit linking proprietary applications with
+the library.  If this is what you want to do, use the GNU Lesser General
+Public License instead of this License.  But first, please read
+<http://www.gnu.org/philosophy/why-not-lgpl.html>.

README.md

@@ -0,0 +1,304 @@
+<div align="center">
+<p>
+   <a align="left" href="https://ultralytics.com/yolov5" target="_blank">
+   <img width="850" src="https://github.com/ultralytics/yolov5/releases/download/v1.0/splash.jpg"></a>
+</p>
+<br>
+<div>
+   <a href="https://github.com/ultralytics/yolov5/actions"><img src="https://github.com/ultralytics/yolov5/workflows/CI%20CPU%20testing/badge.svg" alt="CI CPU testing"></a>
+   <a href="https://zenodo.org/badge/latestdoi/264818686"><img src="https://zenodo.org/badge/264818686.svg" alt="YOLOv5 Citation"></a>
+   <a href="https://hub.docker.com/r/ultralytics/yolov5"><img src="https://img.shields.io/docker/pulls/ultralytics/yolov5?logo=docker" alt="Docker Pulls"></a>
+   <br>
+   <a href="https://colab.research.google.com/github/ultralytics/yolov5/blob/master/tutorial.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"></a>
+   <a href="https://www.kaggle.com/ultralytics/yolov5"><img src="https://kaggle.com/static/images/open-in-kaggle.svg" alt="Open In Kaggle"></a>
+   <a href="https://join.slack.com/t/ultralytics/shared_invite/zt-w29ei8bp-jczz7QYUmDtgo6r6KcMIAg"><img src="https://img.shields.io/badge/Slack-Join_Forum-blue.svg?logo=slack" alt="Join Forum"></a>
+</div>
+
+<br>
+<p>
+YOLOv5 🚀 is a family of object detection architectures and models pretrained on the COCO dataset, and represents <a href="https://ultralytics.com">Ultralytics</a>
+ open-source research into future vision AI methods, incorporating lessons learned and best practices evolved over thousands of hours of research and development.
+</p>
+
+<div align="center">
+   <a href="https://github.com/ultralytics">
+   <img src="https://github.com/ultralytics/yolov5/releases/download/v1.0/logo-social-github.png" width="2%"/>
+   </a>
+   <img width="2%" />
+   <a href="https://www.linkedin.com/company/ultralytics">
+   <img src="https://github.com/ultralytics/yolov5/releases/download/v1.0/logo-social-linkedin.png" width="2%"/>
+   </a>
+   <img width="2%" />
+   <a href="https://twitter.com/ultralytics">
+   <img src="https://github.com/ultralytics/yolov5/releases/download/v1.0/logo-social-twitter.png" width="2%"/>
+   </a>
+   <img width="2%" />
+   <a href="https://www.producthunt.com/@glenn_jocher">
+   <img src="https://github.com/ultralytics/yolov5/releases/download/v1.0/logo-social-producthunt.png" width="2%"/>
+   </a>
+   <img width="2%" />
+   <a href="https://youtube.com/ultralytics">
+   <img src="https://github.com/ultralytics/yolov5/releases/download/v1.0/logo-social-youtube.png" width="2%"/>
+   </a>
+   <img width="2%" />
+   <a href="https://www.facebook.com/ultralytics">
+   <img src="https://github.com/ultralytics/yolov5/releases/download/v1.0/logo-social-facebook.png" width="2%"/>
+   </a>
+   <img width="2%" />
+   <a href="https://www.instagram.com/ultralytics/">
+   <img src="https://github.com/ultralytics/yolov5/releases/download/v1.0/logo-social-instagram.png" width="2%"/>
+   </a>
+</div>
+
+<!--
+<a align="center" href="https://ultralytics.com/yolov5" target="_blank">
+<img width="800" src="https://github.com/ultralytics/yolov5/releases/download/v1.0/banner-api.png"></a>
+-->
+
+</div>
+
+## <div align="center">Documentation</div>
+
+See the [YOLOv5 Docs](https://docs.ultralytics.com) for full documentation on training, testing and deployment.
+
+## <div align="center">Quick Start Examples</div>
+
+<details open>
+<summary>Install</summary>
+
+Clone repo and install [requirements.txt](https://github.com/ultralytics/yolov5/blob/master/requirements.txt) in a
+[**Python>=3.7.0**](https://www.python.org/) environment, including
+[**PyTorch>=1.7**](https://pytorch.org/get-started/locally/).
+
+```bash
+git clone https://github.com/ultralytics/yolov5  # clone
+cd yolov5
+pip install -r requirements.txt  # install
+```
+
+</details>
+
+<details open>
+<summary>Inference</summary>
+
+Inference with YOLOv5 and [PyTorch Hub](https://github.com/ultralytics/yolov5/issues/36)
+. [Models](https://github.com/ultralytics/yolov5/tree/master/models) download automatically from the latest
+YOLOv5 [release](https://github.com/ultralytics/yolov5/releases).
+
+```python
+import torch
+
+# Model
+model = torch.hub.load('ultralytics/yolov5', 'yolov5s')  # or yolov5m, yolov5l, yolov5x, custom
+
+# Images
+img = 'https://ultralytics.com/images/zidane.jpg'  # or file, Path, PIL, OpenCV, numpy, list
+
+# Inference
+results = model(img)
+
+# Results
+results.print()  # or .show(), .save(), .crop(), .pandas(), etc.
+```
+
+</details>
+
+
+
+<details>
+<summary>Inference with detect.py</summary>
+
+`detect.py` runs inference on a variety of sources, downloading [models](https://github.com/ultralytics/yolov5/tree/master/models) automatically from
+the latest YOLOv5 [release](https://github.com/ultralytics/yolov5/releases) and saving results to `runs/detect`.
+
+```bash
+python detect.py --source 0  # webcam
+                          img.jpg  # image
+                          vid.mp4  # video
+                          path/  # directory
+                          path/*.jpg  # glob
+                          'https://youtu.be/Zgi9g1ksQHc'  # YouTube
+                          'rtsp://example.com/media.mp4'  # RTSP, RTMP, HTTP stream
+```
+
+</details>
+
+<details>
+<summary>Training</summary>
+
+The commands below reproduce YOLOv5 [COCO](https://github.com/ultralytics/yolov5/blob/master/data/scripts/get_coco.sh)
+results. [Models](https://github.com/ultralytics/yolov5/tree/master/models)
+and [datasets](https://github.com/ultralytics/yolov5/tree/master/data) download automatically from the latest
+YOLOv5 [release](https://github.com/ultralytics/yolov5/releases). Training times for YOLOv5n/s/m/l/x are
+1/2/4/6/8 days on a V100 GPU ([Multi-GPU](https://github.com/ultralytics/yolov5/issues/475) times faster). Use the
+largest `--batch-size` possible, or pass `--batch-size -1` for
+YOLOv5 [AutoBatch](https://github.com/ultralytics/yolov5/pull/5092). Batch sizes shown for V100-16GB.
+
+```bash
+python train.py --data coco.yaml --cfg yolov5n.yaml --weights '' --batch-size 128
+                                       yolov5s                                64
+                                       yolov5m                                40
+                                       yolov5l                                24
+                                       yolov5x                                16
+```
+
+<img width="800" src="https://user-images.githubusercontent.com/26833433/90222759-949d8800-ddc1-11ea-9fa1-1c97eed2b963.png">
+
+</details>
+
+<details open>
+<summary>Tutorials</summary>
+
+* [Train Custom Data](https://github.com/ultralytics/yolov5/wiki/Train-Custom-Data)&nbsp; 🚀 RECOMMENDED
+* [Tips for Best Training Results](https://github.com/ultralytics/yolov5/wiki/Tips-for-Best-Training-Results)&nbsp; ☘️
+  RECOMMENDED
+* [Weights & Biases Logging](https://github.com/ultralytics/yolov5/issues/1289)&nbsp; 🌟 NEW
+* [Roboflow for Datasets, Labeling, and Active Learning](https://github.com/ultralytics/yolov5/issues/4975)&nbsp; 🌟 NEW
+* [Multi-GPU Training](https://github.com/ultralytics/yolov5/issues/475)
+* [PyTorch Hub](https://github.com/ultralytics/yolov5/issues/36)&nbsp; ⭐ NEW
+* [TFLite, ONNX, CoreML, TensorRT Export](https://github.com/ultralytics/yolov5/issues/251) 🚀
+* [Test-Time Augmentation (TTA)](https://github.com/ultralytics/yolov5/issues/303)
+* [Model Ensembling](https://github.com/ultralytics/yolov5/issues/318)
+* [Model Pruning/Sparsity](https://github.com/ultralytics/yolov5/issues/304)
+* [Hyperparameter Evolution](https://github.com/ultralytics/yolov5/issues/607)
+* [Transfer Learning with Frozen Layers](https://github.com/ultralytics/yolov5/issues/1314)&nbsp; ⭐ NEW
+* [TensorRT Deployment](https://github.com/wang-xinyu/tensorrtx)
+
+</details>
+
+## <div align="center">Environments</div>
+
+Get started in seconds with our verified environments. Click each icon below for details.
+
+<div align="center">
+    <a href="https://colab.research.google.com/github/ultralytics/yolov5/blob/master/tutorial.ipynb">
+        <img src="https://github.com/ultralytics/yolov5/releases/download/v1.0/logo-colab-small.png" width="15%"/>
+    </a>
+    <a href="https://www.kaggle.com/ultralytics/yolov5">
+        <img src="https://github.com/ultralytics/yolov5/releases/download/v1.0/logo-kaggle-small.png" width="15%"/>
+    </a>
+    <a href="https://hub.docker.com/r/ultralytics/yolov5">
+        <img src="https://github.com/ultralytics/yolov5/releases/download/v1.0/logo-docker-small.png" width="15%"/>
+    </a>
+    <a href="https://github.com/ultralytics/yolov5/wiki/AWS-Quickstart">
+        <img src="https://github.com/ultralytics/yolov5/releases/download/v1.0/logo-aws-small.png" width="15%"/>
+    </a>
+    <a href="https://github.com/ultralytics/yolov5/wiki/GCP-Quickstart">
+        <img src="https://github.com/ultralytics/yolov5/releases/download/v1.0/logo-gcp-small.png" width="15%"/>
+    </a>
+</div>
+
+## <div align="center">Integrations</div>
+
+<div align="center">
+    <a href="https://wandb.ai/site?utm_campaign=repo_yolo_readme">
+        <img src="https://github.com/ultralytics/yolov5/releases/download/v1.0/logo-wb-long.png" width="49%"/>
+    </a>
+    <a href="https://roboflow.com/?ref=ultralytics">
+        <img src="https://github.com/ultralytics/yolov5/releases/download/v1.0/logo-roboflow-long.png" width="49%"/>
+    </a>
+</div>
+
+|Weights and Biases|Roboflow ⭐ NEW|
+|:-:|:-:|
+|Automatically track and visualize all your YOLOv5 training runs in the cloud with [Weights & Biases](https://wandb.ai/site?utm_campaign=repo_yolo_readme)|Label and export your custom datasets directly to YOLOv5 for training with [Roboflow](https://roboflow.com/?ref=ultralytics) |
+
+
+<!-- ## <div align="center">Compete and Win</div>
+
+We are super excited about our first-ever Ultralytics YOLOv5 🚀 EXPORT Competition with **$10,000** in cash prizes!
+
+<p align="center">
+  <a href="https://github.com/ultralytics/yolov5/discussions/3213">
+  <img width="850" src="https://github.com/ultralytics/yolov5/releases/download/v1.0/banner-export-competition.png"></a>
+</p> -->
+
+## <div align="center">Why YOLOv5</div>
+
+<p align="left"><img width="800" src="https://user-images.githubusercontent.com/26833433/155040763-93c22a27-347c-4e3c-847a-8094621d3f4e.png"></p>
+<details>
+  <summary>YOLOv5-P5 640 Figure (click to expand)</summary>
+
+<p align="left"><img width="800" src="https://user-images.githubusercontent.com/26833433/155040757-ce0934a3-06a6-43dc-a979-2edbbd69ea0e.png"></p>
+</details>
+<details>
+  <summary>Figure Notes (click to expand)</summary>
+
+* **COCO AP val** denotes mAP@0.5:0.95 metric measured on the 5000-image [COCO val2017](http://cocodataset.org) dataset over various inference sizes from 256 to 1536.
+* **GPU Speed** measures average inference time per image on [COCO val2017](http://cocodataset.org) dataset using a [AWS p3.2xlarge](https://aws.amazon.com/ec2/instance-types/p3/) V100 instance at batch-size 32.
+* **EfficientDet** data from [google/automl](https://github.com/google/automl) at batch size 8.
+* **Reproduce** by `python val.py --task study --data coco.yaml --iou 0.7 --weights yolov5n6.pt yolov5s6.pt yolov5m6.pt yolov5l6.pt yolov5x6.pt`
+</details>
+
+### Pretrained Checkpoints
+
+[assets]: https://github.com/ultralytics/yolov5/releases
+
+[TTA]: https://github.com/ultralytics/yolov5/issues/303
+
+|Model |size<br><sup>(pixels) |mAP<sup>val<br>0.5:0.95 |mAP<sup>val<br>0.5 |Speed<br><sup>CPU b1<br>(ms) |Speed<br><sup>V100 b1<br>(ms) |Speed<br><sup>V100 b32<br>(ms) |params<br><sup>(M) |FLOPs<br><sup>@640 (B)
+|---                    |---  |---    |---    |---    |---    |---    |---    |---
+|[YOLOv5n][assets]      |640  |28.0   |45.7   |**45** |**6.3**|**0.6**|**1.9**|**4.5**
+|[YOLOv5s][assets]      |640  |37.4   |56.8   |98     |6.4    |0.9    |7.2    |16.5
+|[YOLOv5m][assets]      |640  |45.4   |64.1   |224    |8.2    |1.7    |21.2   |49.0
+|[YOLOv5l][assets]      |640  |49.0   |67.3   |430    |10.1   |2.7    |46.5   |109.1
+|[YOLOv5x][assets]      |640  |50.7   |68.9   |766    |12.1   |4.8    |86.7   |205.7
+|                       |     |       |       |       |       |       |       |
+|[YOLOv5n6][assets]     |1280 |36.0   |54.4   |153    |8.1    |2.1    |3.2    |4.6
+|[YOLOv5s6][assets]     |1280 |44.8   |63.7   |385    |8.2    |3.6    |16.8   |12.6
+|[YOLOv5m6][assets]     |1280 |51.3   |69.3   |887    |11.1   |6.8    |35.7   |50.0
+|[YOLOv5l6][assets]     |1280 |53.7   |71.3   |1784   |15.8   |10.5   |76.8   |111.4
+|[YOLOv5x6][assets]<br>+ [TTA][TTA]|1280<br>1536 |55.0<br>**55.8** |72.7<br>**72.7** |3136<br>- |26.2<br>- |19.4<br>- |140.7<br>- |209.8<br>- 
+
+<details>
+  <summary>Table Notes (click to expand)</summary>
+
+* All checkpoints are trained to 300 epochs with default settings. Nano and Small models use [hyp.scratch-low.yaml](https://github.com/ultralytics/yolov5/blob/master/data/hyps/hyp.scratch-low.yaml) hyps, all others use [hyp.scratch-high.yaml](https://github.com/ultralytics/yolov5/blob/master/data/hyps/hyp.scratch-high.yaml).
+* **mAP<sup>val</sup>** values are for single-model single-scale on [COCO val2017](http://cocodataset.org) dataset.<br>Reproduce by `python val.py --data coco.yaml --img 640 --conf 0.001 --iou 0.65`
+* **Speed** averaged over COCO val images using a [AWS p3.2xlarge](https://aws.amazon.com/ec2/instance-types/p3/) instance. NMS times (~1 ms/img) not included.<br>Reproduce by `python val.py --data coco.yaml --img 640 --task speed --batch 1`
+* **TTA** [Test Time Augmentation](https://github.com/ultralytics/yolov5/issues/303) includes reflection and scale augmentations.<br>Reproduce by `python val.py --data coco.yaml --img 1536 --iou 0.7 --augment`
+
+</details>
+
+## <div align="center">Contribute</div>
+
+We love your input! We want to make contributing to YOLOv5 as easy and transparent as possible. Please see our [Contributing Guide](CONTRIBUTING.md) to get started, and fill out the [YOLOv5 Survey](https://ultralytics.com/survey?utm_source=github&utm_medium=social&utm_campaign=Survey) to send us feedback on your experiences. Thank you to all our contributors!
+
+<a href="https://github.com/ultralytics/yolov5/graphs/contributors"><img src="https://opencollective.com/ultralytics/contributors.svg?width=990" /></a>
+
+## <div align="center">Contact</div>
+
+For YOLOv5 bugs and feature requests please visit [GitHub Issues](https://github.com/ultralytics/yolov5/issues). For business inquiries or
+professional support requests please visit [https://ultralytics.com/contact](https://ultralytics.com/contact).
+
+<br>
+
+<div align="center">
+    <a href="https://github.com/ultralytics">
+        <img src="https://github.com/ultralytics/yolov5/releases/download/v1.0/logo-social-github.png" width="3%"/>
+    </a>
+    <img width="3%" />
+    <a href="https://www.linkedin.com/company/ultralytics">
+        <img src="https://github.com/ultralytics/yolov5/releases/download/v1.0/logo-social-linkedin.png" width="3%"/>
+    </a>
+    <img width="3%" />
+    <a href="https://twitter.com/ultralytics">
+        <img src="https://github.com/ultralytics/yolov5/releases/download/v1.0/logo-social-twitter.png" width="3%"/>
+    </a>
+    <img width="3%" />
+    <a href="https://www.producthunt.com/@glenn_jocher">
+    <img src="https://github.com/ultralytics/yolov5/releases/download/v1.0/logo-social-producthunt.png" width="3%"/>
+    </a>
+    <img width="3%" />
+    <a href="https://youtube.com/ultralytics">
+        <img src="https://github.com/ultralytics/yolov5/releases/download/v1.0/logo-social-youtube.png" width="3%"/>
+    </a>
+    <img width="3%" />
+    <a href="https://www.facebook.com/ultralytics">
+        <img src="https://github.com/ultralytics/yolov5/releases/download/v1.0/logo-social-facebook.png" width="3%"/>
+    </a>
+    <img width="3%" />
+    <a href="https://www.instagram.com/ultralytics/">
+        <img src="https://github.com/ultralytics/yolov5/releases/download/v1.0/logo-social-instagram.png" width="3%"/>
+    </a>
+</div>

bash_train_sparity.sh

@@ -0,0 +1,3 @@
+# python train_sparsity.py --st --sr 0.005 --data data/VOC.yaml --cfg models/yolov5s.yaml --weights VOC2007_wm/train/exp5/weights/best.pt --batch-size 16 --epochs 300 --workers 4 --project VOC2007_wm/train_sparity --name exp
+
+python train_sparsity.py --st --sr 0.0001 --weights VOC2007_wm/train/exp5/weights/best.pt --epochs 100 --project VOC2007_wm/train_sparity

blind_watermark/__init__.py

@@ -0,0 +1,5 @@
+from .blind_watermark import WaterMark
+from .bwm_core import WaterMarkCore
+from .att import *
+from .recover import recover_crop
+from .version import __version__, bw_notes

blind_watermark/att.py

@@ -0,0 +1,225 @@
+# coding=utf-8
+
+# attack on the watermark
+import cv2
+import numpy as np
+import warnings
+
+
+def cut_att3(input_filename=None, input_img=None, output_file_name=None, loc_r=None, loc=None, scale=None):
+    # 剪切攻击 + 缩放攻击
+    if input_filename:
+        input_img = cv2.imread(input_filename)
+
+    if loc is None:
+        h, w, _ = input_img.shape
+        x1, y1, x2, y2 = int(w * loc_r[0][0]), int(h * loc_r[0][1]), int(w * loc_r[1][0]), int(h * loc_r[1][1])
+    else:
+        x1, y1, x2, y2 = loc
+
+    # 剪切攻击
+    output_img = input_img[y1:y2, x1:x2].copy()
+
+    # 如果缩放攻击
+    if scale and scale != 1:
+        h, w, _ = output_img.shape
+        output_img = cv2.resize(output_img, dsize=(round(w * scale), round(h * scale)))
+    else:
+        output_img = output_img
+
+    if output_file_name:
+        cv2.imwrite(output_file_name, output_img)
+    return output_img
+
+
+cut_att2 = cut_att3
+
+
+def resize_att(input_filename=None, input_img=None, output_file_name=None, out_shape=(500, 500)):
+    # 缩放攻击：因为攻击和还原都是缩放，所以攻击和还原都调用这个函数
+    if input_filename:
+        input_img = cv2.imread(input_filename)
+    output_img = cv2.resize(input_img, dsize=out_shape)
+    if output_file_name:
+        cv2.imwrite(output_file_name, output_img)
+    return output_img
+
+
+def bright_att(input_filename=None, input_img=None, output_file_name=None, ratio=0.8):
+    # 亮度调整攻击，ratio应当多于0
+    # ratio>1是调得更亮，ratio<1是亮度更暗
+    if input_filename:
+        input_img = cv2.imread(input_filename)
+    output_img = input_img * ratio
+    output_img[output_img > 255] = 255
+    if output_file_name:
+        cv2.imwrite(output_file_name, output_img)
+    return output_img
+
+
+def shelter_att(input_filename=None, input_img=None, output_file_name=None, ratio=0.1, n=3):
+    # 遮挡攻击：遮挡图像中的一部分
+    # n个遮挡块
+    # 每个遮挡块所占比例为ratio
+    if input_filename:
+        output_img = cv2.imread(input_filename)
+    else:
+        output_img = input_img.copy()
+    input_img_shape = output_img.shape
+
+    for i in range(n):
+        tmp = np.random.rand() * (1 - ratio)  # 随机选择一个地方，1-ratio是为了防止溢出
+        start_height, end_height = int(tmp * input_img_shape[0]), int((tmp + ratio) * input_img_shape[0])
+        tmp = np.random.rand() * (1 - ratio)
+        start_width, end_width = int(tmp * input_img_shape[1]), int((tmp + ratio) * input_img_shape[1])
+
+        output_img[start_height:end_height, start_width:end_width, :] = 255
+
+    if output_file_name:
+        cv2.imwrite(output_file_name, output_img)
+    return output_img
+
+
+def salt_pepper_att(input_filename=None, input_img=None, output_file_name=None, ratio=0.01):
+    # 椒盐攻击
+    if input_filename:
+        input_img = cv2.imread(input_filename)
+    input_img_shape = input_img.shape
+    output_img = input_img.copy()
+    for i in range(input_img_shape[0]):
+        for j in range(input_img_shape[1]):
+            if np.random.rand() < ratio:
+                output_img[i, j, :] = 255
+    if output_file_name:
+        cv2.imwrite(output_file_name, output_img)
+    return output_img
+
+
+def rot_att(input_filename=None, input_img=None, output_file_name=None, angle=45):
+    # 旋转攻击
+    if input_filename:
+        input_img = cv2.imread(input_filename)
+    rows, cols, _ = input_img.shape
+    M = cv2.getRotationMatrix2D(center=(cols / 2, rows / 2), angle=angle, scale=1)
+    output_img = cv2.warpAffine(input_img, M, (cols, rows))
+    if output_file_name:
+        cv2.imwrite(output_file_name, output_img)
+    return output_img
+
+
+def cut_att_height(input_filename=None, input_img=None, output_file_name=None, ratio=0.8):
+    warnings.warn('will be deprecated in the future, use att.cut_att2 instead')
+    # 纵向剪切攻击
+    if input_filename:
+        input_img = cv2.imread(input_filename)
+    input_img_shape = input_img.shape
+    height = int(input_img_shape[0] * ratio)
+
+    output_img = input_img[:height, :, :]
+    if output_file_name:
+        cv2.imwrite(output_file_name, output_img)
+    return output_img
+
+
+def cut_att_width(input_filename=None, input_img=None, output_file_name=None, ratio=0.8):
+    warnings.warn('will be deprecated in the future, use att.cut_att2 instead')
+    # 横向裁剪攻击
+    if input_filename:
+        input_img = cv2.imread(input_filename)
+    input_img_shape = input_img.shape
+    width = int(input_img_shape[1] * ratio)
+
+    output_img = input_img[:, :width, :]
+    if output_file_name:
+        cv2.imwrite(output_file_name, output_img)
+    return output_img
+
+
+def cut_att(input_filename=None, output_file_name=None, input_img=None, loc=((0.3, 0.1), (0.7, 0.9)), resize=0.6):
+    warnings.warn('will be deprecated in the future, use att.cut_att2 instead')
+    # 截屏攻击 = 裁剪攻击 + 缩放攻击 + 知道攻击参数（按照参数还原）
+    # 裁剪攻击：其它部分都补0
+    if input_filename:
+        input_img = cv2.imread(input_filename)
+
+    output_img = input_img.copy()
+    shape = output_img.shape
+    x1, y1, x2, y2 = shape[0] * loc[0][0], shape[1] * loc[0][1], shape[0] * loc[1][0], shape[1] * loc[1][1]
+    output_img[:int(x1), :] = 255
+    output_img[int(x2):, :] = 255
+    output_img[:, :int(y1)] = 255
+    output_img[:, int(y2):] = 255
+
+    if resize is not None:
+        # 缩放一次，然后还原
+        output_img = cv2.resize(output_img,
+                                dsize=(int(shape[1] * resize), int(shape[0] * resize))
+                                )
+
+        output_img = cv2.resize(output_img, dsize=(int(shape[1]), int(shape[0])))
+
+    if output_file_name is not None:
+        cv2.imwrite(output_file_name, output_img)
+    return output_img
+
+
+# def cut_att2(input_filename=None, input_img=None, output_file_name=None, loc_r=((0.3, 0.1), (0.9, 0.9)), scale=1.1):
+#     # 截屏攻击 = 剪切攻击 + 缩放攻击 + 不知道攻击参数
+#     if input_filename:
+#         input_img = cv2.imread(input_filename)
+#     h, w, _ = input_img.shape
+#     x1, y1, x2, y2 = int(w * loc_r[0][0]), int(h * loc_r[0][1]), int(w * loc_r[1][0]), int(h * loc_r[1][1])
+#
+#     output_img = cut_att3(input_img=input_img, output_file_name=output_file_name,
+#                           loc=(x1, y1, x2, y2), scale=scale)
+#     return output_img, (x1, y1, x2, y2)
+
+def anti_cut_att_old(input_filename, output_file_name, origin_shape):
+    warnings.warn('will be deprecated in the future')
+    # 反裁剪攻击：复制一块范围，然后补全
+    # origin_shape 分辨率与约定理解的是颠倒的，约定的是列数*行数
+    input_img = cv2.imread(input_filename)
+    output_img = input_img.copy()
+    output_img_shape = output_img.shape
+    if output_img_shape[0] > origin_shape[0] or output_img_shape[0] > origin_shape[0]:
+        print('裁剪打击后的图片，不可能比原始图片大，检查一下')
+        return
+
+    # 还原纵向打击
+    while output_img_shape[0] < origin_shape[0]:
+        output_img = np.concatenate([output_img, output_img[:origin_shape[0] - output_img_shape[0], :, :]], axis=0)
+        output_img_shape = output_img.shape
+    while output_img_shape[1] < origin_shape[1]:
+        output_img = np.concatenate([output_img, output_img[:, :origin_shape[1] - output_img_shape[1], :]], axis=1)
+        output_img_shape = output_img.shape
+
+    cv2.imwrite(output_file_name, output_img)
+
+
+def anti_cut_att(input_filename=None, input_img=None, output_file_name=None, origin_shape=None):
+    warnings.warn('will be deprecated in the future, use att.cut_att2 instead')
+    # 反裁剪攻击：补0
+    # origin_shape 分辨率与约定理解的是颠倒的，约定的是列数*行数
+    if input_filename:
+        input_img = cv2.imread(input_filename)
+    output_img = input_img.copy()
+    output_img_shape = output_img.shape
+    if output_img_shape[0] > origin_shape[0] or output_img_shape[0] > origin_shape[0]:
+        print('裁剪打击后的图片，不可能比原始图片大，检查一下')
+        return
+
+    # 还原纵向打击
+    if output_img_shape[0] < origin_shape[0]:
+        output_img = np.concatenate(
+            [output_img, 255 * np.ones((origin_shape[0] - output_img_shape[0], output_img_shape[1], 3))]
+            , axis=0)
+        output_img_shape = output_img.shape
+
+    if output_img_shape[1] < origin_shape[1]:
+        output_img = np.concatenate(
+            [output_img, 255 * np.ones((output_img_shape[0], origin_shape[1] - output_img_shape[1], 3))]
+            , axis=1)
+
+    if output_file_name:
+        cv2.imwrite(output_file_name, output_img)
+    return output_img

blind_watermark/blind_watermark.py

@@ -0,0 +1,109 @@
+#!/usr/bin/env python3
+# coding=utf-8
+# @Time    : 2020/8/13
+# @Author  : github.com/guofei9987
+import warnings
+
+import numpy as np
+import cv2
+
+from .bwm_core import WaterMarkCore
+from .version import bw_notes
+
+
+class WaterMark:
+    def __init__(self, password_wm=1, password_img=1, block_shape=(4, 4), mode='common', processes=None):
+        bw_notes.print_notes()
+
+        self.bwm_core = WaterMarkCore(password_img=password_img, mode=mode, processes=processes)
+
+        self.password_wm = password_wm
+
+        self.wm_bit = None
+        self.wm_size = 0
+
+    def read_img(self, filename=None, img=None):
+        if img is None:
+            # 从文件读入图片
+            img = cv2.imread(filename, flags=cv2.IMREAD_UNCHANGED)
+            assert img is not None, "image file '{filename}' not read".format(filename=filename)
+
+        self.bwm_core.read_img_arr(img=img)
+        return img
+
+    def read_wm(self, wm_content, mode='img'):
+        assert mode in ('img', 'str', 'bit'), "mode in ('img','str','bit')"
+        if mode == 'img':
+            wm = cv2.imread(filename=wm_content, flags=cv2.IMREAD_GRAYSCALE)
+            assert wm is not None, 'file "{filename}" not read'.format(filename=wm_content)
+
+            # 读入图片格式的水印，并转为一维 bit 格式，抛弃灰度级别
+            self.wm_bit = wm.flatten() > 128
+
+        elif mode == 'str':
+            byte = bin(int(wm_content.encode('utf-8').hex(), base=16))[2:]
+            self.wm_bit = (np.array(list(byte)) == '1')
+        else:
+            self.wm_bit = np.array(wm_content)
+
+        self.wm_size = self.wm_bit.size
+
+        # 水印加密:
+        np.random.RandomState(self.password_wm).shuffle(self.wm_bit)
+
+        self.bwm_core.read_wm(self.wm_bit)
+
+    def embed(self, filename=None, compression_ratio=None):
+        '''
+        :param filename: string
+            Save the image file as filename
+        :param compression_ratio: int or None
+            If compression_ratio = None, do not compression,
+            If compression_ratio is integer between 0 and 100, the smaller, the output file is smaller.
+        :return:
+        '''
+        embed_img = self.bwm_core.embed()
+        if filename is not None:
+            if compression_ratio is None:
+                cv2.imwrite(filename=filename, img=embed_img)
+            elif filename.endswith('.jpg'):
+                cv2.imwrite(filename=filename, img=embed_img, params=[cv2.IMWRITE_JPEG_QUALITY, compression_ratio])
+            elif filename.endswith('.png'):
+                cv2.imwrite(filename=filename, img=embed_img, params=[cv2.IMWRITE_PNG_COMPRESSION, compression_ratio])
+            else:
+                cv2.imwrite(filename=filename, img=embed_img)
+        return embed_img
+
+    def extract_decrypt(self, wm_avg):
+        wm_index = np.arange(self.wm_size)
+        np.random.RandomState(self.password_wm).shuffle(wm_index)
+        wm_avg[wm_index] = wm_avg.copy()
+        return wm_avg
+
+    def extract(self, filename=None, embed_img=None, wm_shape=None, out_wm_name=None, mode='img'):
+        assert wm_shape is not None, 'wm_shape needed'
+
+        if filename is not None:
+            embed_img = cv2.imread(filename, flags=cv2.IMREAD_COLOR)
+            assert embed_img is not None, "{filename} not read".format(filename=filename)
+
+        self.wm_size = np.array(wm_shape).prod()
+
+        if mode in ('str', 'bit'):
+            wm_avg = self.bwm_core.extract_with_kmeans(img=embed_img, wm_shape=wm_shape)
+        else:
+            wm_avg = self.bwm_core.extract(img=embed_img, wm_shape=wm_shape)
+
+        # 解密：
+        wm = self.extract_decrypt(wm_avg=wm_avg)
+
+        # 转化为指定格式：
+        if mode == 'img':
+            wm = 255 * wm.reshape(wm_shape[0], wm_shape[1])
+            cv2.imwrite(out_wm_name, wm)
+        elif mode == 'str':
+            byte = ''.join(str((i >= 0.5) * 1) for i in wm)
+            print("Byte value:", byte)
+            wm = bytes.fromhex(hex(int(byte, base=2))[2:]).decode('utf-8', errors='replace')
+
+        return wm

blind_watermark/bwm_core.py

@@ -0,0 +1,266 @@
+#!/usr/bin/env python3
+# coding=utf-8
+# @Time    : 2021/12/17
+# @Author  : github.com/guofei9987
+import numpy as np
+from numpy.linalg import svd
+import copy
+import cv2
+from cv2 import dct, idct
+from pywt import dwt2, idwt2
+from .pool import AutoPool
+
+
+class WaterMarkCore:
+    def __init__(self, password_img=1, mode='common', processes=None):
+        self.block_shape = np.array([4, 4])
+        self.password_img = password_img
+        self.d1, self.d2 = 36, 20  # d1/d2 越大鲁棒性越强,但输出图片的失真越大
+
+        # init data
+        self.img, self.img_YUV = None, None  # self.img 是原图，self.img_YUV 对像素做了加白偶数化
+        self.ca, self.hvd, = [np.array([])] * 3, [np.array([])] * 3  # 每个通道 dct 的结果
+        self.ca_block = [np.array([])] * 3  # 每个 channel 存一个四维 array，代表四维分块后的结果
+        self.ca_part = [np.array([])] * 3  # 四维分块后，有时因不整除而少一部分，self.ca_part 是少这一部分的 self.ca
+
+        self.wm_size, self.block_num = 0, 0  # 水印的长度，原图片可插入信息的个数
+        self.pool = AutoPool(mode=mode, processes=processes)
+
+        self.fast_mode = False
+        self.alpha = None  # 用于处理透明图
+
+    def init_block_index(self):
+        self.block_num = self.ca_block_shape[0] * self.ca_block_shape[1]
+        assert self.wm_size < self.block_num, IndexError(
+            '最多可嵌入{}kb信息，多于水印的{}kb信息，溢出'.format(self.block_num / 1000, self.wm_size / 1000))
+        # self.part_shape 是取整后的ca二维大小,用于嵌入时忽略右边和下面对不齐的细条部分。
+        self.part_shape = self.ca_block_shape[:2] * self.block_shape
+        self.block_index = [(i, j) for i in range(self.ca_block_shape[0]) for j in range(self.ca_block_shape[1])]
+
+    # def read_img_arr(self, img):
+    #     # 处理透明图
+    #     self.alpha = None
+    #     if img.shape[2] == 4:
+    #         if img[:, :, 3].min() < 255:
+    #             self.alpha = img[:, :, 3]
+    #             img = img[:, :, :3]
+
+    #     # 读入图片->YUV化->加白边使像素变偶数->四维分块
+    #     self.img = img.astype(np.float32)
+    #     self.img_shape = self.img.shape[:2]
+
+    #     # 如果不是偶数，那么补上白边，Y（明亮度）UV（颜色）
+    #     self.img_YUV = cv2.copyMakeBorder(cv2.cvtColor(self.img, cv2.COLOR_BGR2YUV),
+    #                                       0, self.img.shape[0] % 2, 0, self.img.shape[1] % 2,
+    #                                       cv2.BORDER_CONSTANT, value=(0, 0, 0))
+
+    #     self.ca_shape = [(i + 1) // 2 for i in self.img_shape]
+
+    #     self.ca_block_shape = (self.ca_shape[0] // self.block_shape[0], self.ca_shape[1] // self.block_shape[1],
+    #                            self.block_shape[0], self.block_shape[1])
+    #     strides = 4 * np.array([self.ca_shape[1] * self.block_shape[0], self.block_shape[1], self.ca_shape[1], 1])
+
+    #     for channel in range(3):
+    #         self.ca[channel], self.hvd[channel] = dwt2(self.img_YUV[:, :, channel], 'haar')
+    #         # 转为4维度
+    #         self.ca_block[channel] = np.lib.stride_tricks.as_strided(self.ca[channel].astype(np.float32),
+    #                                                                  self.ca_block_shape, strides)
+    def read_img_arr(self, img):
+        # 处理透明图
+        self.alpha = None
+        if len(img.shape) == 3 and img.shape[2] == 4:
+            if img[:, :, 3].min() < 255:
+                self.alpha = img[:, :, 3]
+                img = img[:, :, :3]
+        
+        # 读入图片->YUV化->加白边使像素变偶数->四维分块
+        self.img = img.astype(np.float32)
+        self.img_shape = self.img.shape[:2]
+
+        # 如果不是偶数，那么补上白边，Y（明亮度）UV（颜色）
+        if len(self.img.shape) == 3:  # RGB图像
+            self.img_YUV = cv2.copyMakeBorder(cv2.cvtColor(self.img, cv2.COLOR_BGR2YUV),
+                                            0, self.img.shape[0] % 2, 0, self.img.shape[1] % 2,
+                                            cv2.BORDER_CONSTANT, value=(0, 0, 0))
+        elif len(self.img.shape) == 2:  # 灰度图像
+            self.img_YUV = cv2.copyMakeBorder(cv2.cvtColor(self.img, cv2.COLOR_GRAY2BGR),
+                                            0, self.img.shape[0] % 2, 0, self.img.shape[1] % 2,
+                                            cv2.BORDER_CONSTANT, value=(0, 0, 0))
+        
+        self.ca_shape = [(i + 1) // 2 for i in self.img_shape]
+
+        self.ca_block_shape = (self.ca_shape[0] // self.block_shape[0], self.ca_shape[1] // self.block_shape[1],
+                            self.block_shape[0], self.block_shape[1])
+        strides = 4 * np.array([self.ca_shape[1] * self.block_shape[0], self.block_shape[1], self.ca_shape[1], 1])
+
+        for channel in range(3):
+            self.ca[channel], self.hvd[channel] = dwt2(self.img_YUV[:, :, channel], 'haar')
+            # 转为4维度
+            self.ca_block[channel] = np.lib.stride_tricks.as_strided(self.ca[channel].astype(np.float32),
+                                                                    self.ca_block_shape, strides)
+
+
+    def read_wm(self, wm_bit):
+        self.wm_bit = wm_bit
+        self.wm_size = wm_bit.size
+
+    def block_add_wm(self, arg):
+        if self.fast_mode:
+            return self.block_add_wm_fast(arg)
+        else:
+            return self.block_add_wm_slow(arg)
+
+    def block_add_wm_slow(self, arg):
+        block, shuffler, i = arg
+        # dct->(flatten->加密->逆flatten)->svd->打水印->逆svd->(flatten->解密->逆flatten)->逆dct
+        wm_1 = self.wm_bit[i % self.wm_size]
+        block_dct = dct(block)
+
+        # 加密（打乱顺序）
+        block_dct_shuffled = block_dct.flatten()[shuffler].reshape(self.block_shape)
+        u, s, v = svd(block_dct_shuffled)
+        s[0] = (s[0] // self.d1 + 1 / 4 + 1 / 2 * wm_1) * self.d1
+        if self.d2:
+            s[1] = (s[1] // self.d2 + 1 / 4 + 1 / 2 * wm_1) * self.d2
+
+        block_dct_flatten = np.dot(u, np.dot(np.diag(s), v)).flatten()
+        block_dct_flatten[shuffler] = block_dct_flatten.copy()
+        return idct(block_dct_flatten.reshape(self.block_shape))
+
+    def block_add_wm_fast(self, arg):
+        # dct->svd->打水印->逆svd->逆dct
+        block, shuffler, i = arg
+        wm_1 = self.wm_bit[i % self.wm_size]
+
+        u, s, v = svd(dct(block))
+        s[0] = (s[0] // self.d1 + 1 / 4 + 1 / 2 * wm_1) * self.d1
+
+        return idct(np.dot(u, np.dot(np.diag(s), v)))
+
+    def embed(self):
+        self.init_block_index()
+
+        embed_ca = copy.deepcopy(self.ca)
+        embed_YUV = [np.array([])] * 3
+
+        self.idx_shuffle = random_strategy1(self.password_img, self.block_num,
+                                            self.block_shape[0] * self.block_shape[1])
+        for channel in range(3):
+            tmp = self.pool.map(self.block_add_wm,
+                                [(self.ca_block[channel][self.block_index[i]], self.idx_shuffle[i], i)
+                                 for i in range(self.block_num)])
+
+            for i in range(self.block_num):
+                self.ca_block[channel][self.block_index[i]] = tmp[i]
+
+            # 4维分块变回2维
+            self.ca_part[channel] = np.concatenate(np.concatenate(self.ca_block[channel], 1), 1)
+            # 4维分块时右边和下边不能整除的长条保留，其余是主体部分，换成 embed 之后的频域的数据
+            embed_ca[channel][:self.part_shape[0], :self.part_shape[1]] = self.ca_part[channel]
+            # 逆变换回去
+            embed_YUV[channel] = idwt2((embed_ca[channel], self.hvd[channel]), "haar")
+
+        # 合并3通道
+        embed_img_YUV = np.stack(embed_YUV, axis=2)
+        # 之前如果不是2的整数，增加了白边，这里去除掉
+        embed_img_YUV = embed_img_YUV[:self.img_shape[0], :self.img_shape[1]]
+        embed_img = cv2.cvtColor(embed_img_YUV, cv2.COLOR_YUV2BGR)
+        embed_img = np.clip(embed_img, a_min=0, a_max=255)
+
+        if self.alpha is not None:
+            embed_img = cv2.merge([embed_img.astype(np.uint8), self.alpha])
+        return embed_img
+
+    def block_get_wm(self, args):
+        if self.fast_mode:
+            return self.block_get_wm_fast(args)
+        else:
+            return self.block_get_wm_slow(args)
+
+    def block_get_wm_slow(self, args):
+        block, shuffler = args
+        # dct->flatten->加密->逆flatten->svd->解水印
+        block_dct_shuffled = dct(block).flatten()[shuffler].reshape(self.block_shape)
+
+        u, s, v = svd(block_dct_shuffled)
+        wm = (s[0] % self.d1 > self.d1 / 2) * 1
+        if self.d2:
+            tmp = (s[1] % self.d2 > self.d2 / 2) * 1
+            wm = (wm * 3 + tmp * 1) / 4
+        return wm
+
+    def block_get_wm_fast(self, args):
+        block, shuffler = args
+        # dct->svd->解水印
+        u, s, v = svd(dct(block))
+        wm = (s[0] % self.d1 > self.d1 / 2) * 1
+
+        return wm
+
+    def extract_raw(self, img):
+        # 每个分块提取 1 bit 信息
+        self.read_img_arr(img=img)
+        self.init_block_index()
+
+        wm_block_bit = np.zeros(shape=(3, self.block_num))  # 3个channel，length 个分块提取的水印，全都记录下来
+
+        self.idx_shuffle = random_strategy1(seed=self.password_img,
+                                            size=self.block_num,
+                                            block_shape=self.block_shape[0] * self.block_shape[1],  # 16
+                                            )
+        for channel in range(3):
+            wm_block_bit[channel, :] = self.pool.map(self.block_get_wm,
+                                                     [(self.ca_block[channel][self.block_index[i]], self.idx_shuffle[i])
+                                                      for i in range(self.block_num)])
+        return wm_block_bit
+
+    def extract_avg(self, wm_block_bit):
+        # 对循环嵌入+3个 channel 求平均
+        wm_avg = np.zeros(shape=self.wm_size)
+        for i in range(self.wm_size):
+            wm_avg[i] = wm_block_bit[:, i::self.wm_size].mean()
+        return wm_avg
+
+    def extract(self, img, wm_shape):
+        self.wm_size = np.array(wm_shape).prod()
+
+        # 提取每个分块埋入的 bit：
+        wm_block_bit = self.extract_raw(img=img)
+        # 做平均：
+        wm_avg = self.extract_avg(wm_block_bit)
+        return wm_avg
+
+    def extract_with_kmeans(self, img, wm_shape):
+        wm_avg = self.extract(img=img, wm_shape=wm_shape)
+
+        return one_dim_kmeans(wm_avg)
+
+
+def one_dim_kmeans(inputs):
+    threshold = 0
+    e_tol = 10 ** (-6)
+    center = [inputs.min(), inputs.max()]  # 1. 初始化中心点
+    for i in range(300):
+        threshold = (center[0] + center[1]) / 2
+        is_class01 = inputs > threshold  # 2. 检查所有点与这k个点之间的距离，每个点归类到最近的中心
+        center = [inputs[~is_class01].mean(), inputs[is_class01].mean()]  # 3. 重新找中心点
+        if np.abs((center[0] + center[1]) / 2 - threshold) < e_tol:  # 4. 停止条件
+            threshold = (center[0] + center[1]) / 2
+            break
+
+    is_class01 = inputs > threshold
+    return is_class01
+
+
+def random_strategy1(seed, size, block_shape):
+    return np.random.RandomState(seed) \
+        .random(size=(size, block_shape)) \
+        .argsort(axis=1)
+
+
+def random_strategy2(seed, size, block_shape):
+    one_line = np.random.RandomState(seed) \
+        .random(size=(1, block_shape)) \
+        .argsort(axis=1)
+
+    return np.repeat(one_line, repeats=size, axis=0)

blind_watermark/cli_tools.py

@@ -0,0 +1,53 @@
+from optparse import OptionParser
+from .blind_watermark import WaterMark
+
+usage1 = 'blind_watermark --embed --pwd 1234 image.jpg "watermark text" embed.png'
+usage2 = 'blind_watermark --extract --pwd 1234 --wm_shape 111 embed.png'
+optParser = OptionParser(usage=usage1 + '\n' + usage2)
+
+optParser.add_option('--embed', dest='work_mode', action='store_const', const='embed'
+                     , help='Embed watermark into images')
+optParser.add_option('--extract', dest='work_mode', action='store_const', const='extract'
+                     , help='Extract watermark from images')
+
+optParser.add_option('-p', '--pwd', dest='password', help='password, like 1234')
+optParser.add_option('--wm_shape', dest='wm_shape', help='Watermark shape, like 120')
+
+(opts, args) = optParser.parse_args()
+
+
+def main():
+    bwm1 = WaterMark(password_img=int(opts.password))
+    if opts.work_mode == 'embed':
+        if not len(args) == 3:
+            print('Error! Usage: ')
+            print(usage1)
+            return
+        else:
+            bwm1.read_img(args[0])
+            bwm1.read_wm(args[1], mode='str')
+            bwm1.embed(args[2])
+            print('Embed succeed! to file ', args[2])
+            print('Put down watermark size:', len(bwm1.wm_bit))
+
+    if opts.work_mode == 'extract':
+        if not len(args) == 1:
+            print('Error! Usage: ')
+            print(usage2)
+            return
+
+        else:
+            wm_str = bwm1.extract(filename=args[0], wm_shape=int(opts.wm_shape), mode='str')
+            print('Extract succeed! watermark is:')
+            print(wm_str)
+
+
+'''
+python -m blind_watermark.cli_tools --embed --pwd 1234 examples/pic/ori_img.jpeg "watermark text" examples/output/embedded.png
+python -m blind_watermark.cli_tools --extract --pwd 1234 --wm_shape 111 examples/output/embedded.png
+
+
+cd examples
+blind_watermark --embed --pwd 1234 examples/pic/ori_img.jpeg "watermark text" examples/output/embedded.png
+blind_watermark --extract --pwd 1234 --wm_shape 111 examples/output/embedded.png
+'''

blind_watermark/pool.py

@@ -0,0 +1,38 @@
+import sys
+import multiprocessing
+import warnings
+
+if sys.platform != 'win32':
+    multiprocessing.set_start_method('fork')
+
+
+class CommonPool(object):
+    def map(self, func, args):
+        return list(map(func, args))
+
+
+class AutoPool(object):
+    def __init__(self, mode, processes):
+
+        if mode == 'multiprocessing' and sys.platform == 'win32':
+            warnings.warn('multiprocessing not support in windows, turning to multithreading')
+            mode = 'multithreading'
+
+        self.mode = mode
+        self.processes = processes
+
+        if mode == 'vectorization':
+            pass
+        elif mode == 'cached':
+            pass
+        elif mode == 'multithreading':
+            from multiprocessing.dummy import Pool as ThreadPool
+            self.pool = ThreadPool(processes=processes)
+        elif mode == 'multiprocessing':
+            from multiprocessing import Pool
+            self.pool = Pool(processes=processes)
+        else:  # common
+            self.pool = CommonPool()
+
+    def map(self, func, args):
+        return self.pool.map(func, args)

blind_watermark/recover.py

@@ -0,0 +1,100 @@
+import cv2
+import numpy as np
+
+import functools
+
+
+# 一个帮助缓存化加速的类，引入事实上的全局变量
+class MyValues:
+    def __init__(self):
+        self.idx = 0
+        self.image, self.template = None, None
+
+    def set_val(self, image, template):
+        self.idx += 1
+        self.image, self.template = image, template
+
+
+my_value = MyValues()
+
+
+@functools.lru_cache(maxsize=None, typed=False)
+def match_template(w, h, idx):
+    image, template = my_value.image, my_value.template
+    resized = cv2.resize(template, dsize=(w, h))
+    scores = cv2.matchTemplate(image, resized, cv2.TM_CCOEFF_NORMED)
+    ind = np.unravel_index(np.argmax(scores, axis=None), scores.shape)
+    return ind, scores[ind]
+
+
+def match_template_by_scale(scale):
+    image, template = my_value.image, my_value.template
+    w, h = round(template.shape[1] * scale), round(template.shape[0] * scale)
+    ind, score = match_template(w, h, idx=my_value.idx)
+    return ind, score, scale
+
+
+def search_template(scale=(0.5, 2), search_num=200):
+    image, template = my_value.image, my_value.template
+    # 局部暴力搜索算法，寻找最优的scale
+    tmp = []
+    min_scale, max_scale = scale
+
+    max_scale = min(max_scale, image.shape[0] / template.shape[0], image.shape[1] / template.shape[1])
+
+    max_idx = 0
+
+    for i in range(2):
+        for scale in np.linspace(min_scale, max_scale, search_num):
+            ind, score, scale = match_template_by_scale(scale)
+            tmp.append([ind, score, scale])
+
+        # 寻找最佳
+        max_idx = 0
+        max_score = 0
+        for idx, (ind, score, scale) in enumerate(tmp):
+            if score > max_score:
+                max_idx, max_score = idx, score
+
+        min_scale, max_scale = tmp[max(0, max_idx - 1)][2], tmp[min(len(tmp) - 1, max_idx + 1)][2]
+
+        search_num = 2 * int((max_scale - min_scale) * max(template.shape[1], template.shape[0])) + 1
+
+    return tmp[max_idx]
+
+
+def estimate_crop_parameters(original_file=None, template_file=None, ori_img=None, tem_img=None
+                             , scale=(0.5, 2), search_num=200):
+    # 推测攻击后的图片，在原图片中的位置、大小
+    if template_file:
+        tem_img = cv2.imread(template_file, cv2.IMREAD_GRAYSCALE)  # template image
+    if original_file:
+        ori_img = cv2.imread(original_file, cv2.IMREAD_GRAYSCALE)  # image
+
+    if scale[0] == scale[1] == 1:
+        # 不缩放
+        scale_infer = 1
+        scores = cv2.matchTemplate(ori_img, tem_img, cv2.TM_CCOEFF_NORMED)
+        ind = np.unravel_index(np.argmax(scores, axis=None), scores.shape)
+        ind, score = ind, scores[ind]
+    else:
+        my_value.set_val(image=ori_img, template=tem_img)
+        ind, score, scale_infer = search_template(scale=scale, search_num=search_num)
+    w, h = int(tem_img.shape[1] * scale_infer), int(tem_img.shape[0] * scale_infer)
+    x1, y1, x2, y2 = ind[1], ind[0], ind[1] + w, ind[0] + h
+    return (x1, y1, x2, y2), ori_img.shape, score, scale_infer
+
+
+def recover_crop(template_file=None, tem_img=None, output_file_name=None, loc=None, image_o_shape=None):
+    if template_file:
+        tem_img = cv2.imread(template_file)  # template image
+
+    (x1, y1, x2, y2) = loc
+
+    img_recovered = np.zeros((image_o_shape[0], image_o_shape[1], 3))
+
+    img_recovered[y1:y2, x1:x2, :] = cv2.resize(tem_img, dsize=(x2 - x1, y2 - y1))
+
+    if output_file_name:
+        cv2.imwrite(output_file_name, img_recovered)
+    return img_recovered

blind_watermark/requirements.txt

@@ -0,0 +1 @@
+blind-watermark

blind_watermark/version.py

@@ -0,0 +1,22 @@
+__version__ = '0.4.4'
+
+
+class Notes:
+    def __init__(self):
+        self.show = True
+
+    def print_notes(self):
+        if self.show:
+            print(f'''
+Welcome to use blind-watermark, version = {__version__}
+Make sure the version is the same when encode and decode
+Your star means a lot: https://github.com/guofei9987/blind_watermark
+This message only show once. To close it: `blind_watermark.bw_notes.close()`
+            ''')
+            self.close()
+
+    def close(self):
+        self.show = False
+
+
+bw_notes = Notes()

data/Argoverse.yaml

@@ -0,0 +1,67 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+# Argoverse-HD dataset (ring-front-center camera) http://www.cs.cmu.edu/~mengtial/proj/streaming/ by Argo AI
+# Example usage: python train.py --data Argoverse.yaml
+# parent
+# ├── yolov5
+# └── datasets
+#     └── Argoverse  ← downloads here
+
+
+# Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..]
+path: ../datasets/Argoverse  # dataset root dir
+train: Argoverse-1.1/images/train/  # train images (relative to 'path') 39384 images
+val: Argoverse-1.1/images/val/  # val images (relative to 'path') 15062 images
+test: Argoverse-1.1/images/test/  # test images (optional) https://eval.ai/web/challenges/challenge-page/800/overview
+
+# Classes
+nc: 8  # number of classes
+names: ['person',  'bicycle',  'car',  'motorcycle',  'bus',  'truck',  'traffic_light',  'stop_sign']  # class names
+
+
+# Download script/URL (optional) ---------------------------------------------------------------------------------------
+download: |
+  import json
+
+  from tqdm import tqdm
+  from utils.general import download, Path
+
+
+  def argoverse2yolo(set):
+      labels = {}
+      a = json.load(open(set, "rb"))
+      for annot in tqdm(a['annotations'], desc=f"Converting {set} to YOLOv5 format..."):
+          img_id = annot['image_id']
+          img_name = a['images'][img_id]['name']
+          img_label_name = img_name[:-3] + "txt"
+
+          cls = annot['category_id']  # instance class id
+          x_center, y_center, width, height = annot['bbox']
+          x_center = (x_center + width / 2) / 1920.0  # offset and scale
+          y_center = (y_center + height / 2) / 1200.0  # offset and scale
+          width /= 1920.0  # scale
+          height /= 1200.0  # scale
+
+          img_dir = set.parents[2] / 'Argoverse-1.1' / 'labels' / a['seq_dirs'][a['images'][annot['image_id']]['sid']]
+          if not img_dir.exists():
+              img_dir.mkdir(parents=True, exist_ok=True)
+
+          k = str(img_dir / img_label_name)
+          if k not in labels:
+              labels[k] = []
+          labels[k].append(f"{cls} {x_center} {y_center} {width} {height}\n")
+
+      for k in labels:
+          with open(k, "w") as f:
+              f.writelines(labels[k])
+
+
+  # Download
+  dir = Path('../datasets/Argoverse')  # dataset root dir
+  urls = ['https://argoverse-hd.s3.us-east-2.amazonaws.com/Argoverse-HD-Full.zip']
+  download(urls, dir=dir, delete=False)
+
+  # Convert
+  annotations_dir = 'Argoverse-HD/annotations/'
+  (dir / 'Argoverse-1.1' / 'tracking').rename(dir / 'Argoverse-1.1' / 'images')  # rename 'tracking' to 'images'
+  for d in "train.json", "val.json":
+      argoverse2yolo(dir / annotations_dir / d)  # convert VisDrone annotations to YOLO labels

data/GlobalWheat2020.yaml

@@ -0,0 +1,53 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+# Global Wheat 2020 dataset http://www.global-wheat.com/ by University of Saskatchewan
+# Example usage: python train.py --data GlobalWheat2020.yaml
+# parent
+# ├── yolov5
+# └── datasets
+#     └── GlobalWheat2020  ← downloads here
+
+
+# Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..]
+path: ../datasets/GlobalWheat2020  # dataset root dir
+train: # train images (relative to 'path') 3422 images
+  - images/arvalis_1
+  - images/arvalis_2
+  - images/arvalis_3
+  - images/ethz_1
+  - images/rres_1
+  - images/inrae_1
+  - images/usask_1
+val: # val images (relative to 'path') 748 images (WARNING: train set contains ethz_1)
+  - images/ethz_1
+test: # test images (optional) 1276 images
+  - images/utokyo_1
+  - images/utokyo_2
+  - images/nau_1
+  - images/uq_1
+
+# Classes
+nc: 1  # number of classes
+names: ['wheat_head']  # class names
+
+
+# Download script/URL (optional) ---------------------------------------------------------------------------------------
+download: |
+  from utils.general import download, Path
+
+  # Download
+  dir = Path(yaml['path'])  # dataset root dir
+  urls = ['https://zenodo.org/record/4298502/files/global-wheat-codalab-official.zip',
+          'https://github.com/ultralytics/yolov5/releases/download/v1.0/GlobalWheat2020_labels.zip']
+  download(urls, dir=dir)
+
+  # Make Directories
+  for p in 'annotations', 'images', 'labels':
+      (dir / p).mkdir(parents=True, exist_ok=True)
+
+  # Move
+  for p in 'arvalis_1', 'arvalis_2', 'arvalis_3', 'ethz_1', 'rres_1', 'inrae_1', 'usask_1', \
+           'utokyo_1', 'utokyo_2', 'nau_1', 'uq_1':
+      (dir / p).rename(dir / 'images' / p)  # move to /images
+      f = (dir / p).with_suffix('.json')  # json file
+      if f.exists():
+          f.rename((dir / 'annotations' / p).with_suffix('.json'))  # move to /annotations

data/Objects365.yaml

@@ -0,0 +1,112 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+# Objects365 dataset https://www.objects365.org/ by Megvii
+# Example usage: python train.py --data Objects365.yaml
+# parent
+# ├── yolov5
+# └── datasets
+#     └── Objects365  ← downloads here
+
+
+# Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..]
+path: ../datasets/Objects365  # dataset root dir
+train: images/train  # train images (relative to 'path') 1742289 images
+val: images/val # val images (relative to 'path') 80000 images
+test:  # test images (optional)
+
+# Classes
+nc: 365  # number of classes
+names: ['Person', 'Sneakers', 'Chair', 'Other Shoes', 'Hat', 'Car', 'Lamp', 'Glasses', 'Bottle', 'Desk', 'Cup',
+        'Street Lights', 'Cabinet/shelf', 'Handbag/Satchel', 'Bracelet', 'Plate', 'Picture/Frame', 'Helmet', 'Book',
+        'Gloves', 'Storage box', 'Boat', 'Leather Shoes', 'Flower', 'Bench', 'Potted Plant', 'Bowl/Basin', 'Flag',
+        'Pillow', 'Boots', 'Vase', 'Microphone', 'Necklace', 'Ring', 'SUV', 'Wine Glass', 'Belt', 'Monitor/TV',
+        'Backpack', 'Umbrella', 'Traffic Light', 'Speaker', 'Watch', 'Tie', 'Trash bin Can', 'Slippers', 'Bicycle',
+        'Stool', 'Barrel/bucket', 'Van', 'Couch', 'Sandals', 'Basket', 'Drum', 'Pen/Pencil', 'Bus', 'Wild Bird',
+        'High Heels', 'Motorcycle', 'Guitar', 'Carpet', 'Cell Phone', 'Bread', 'Camera', 'Canned', 'Truck',
+        'Traffic cone', 'Cymbal', 'Lifesaver', 'Towel', 'Stuffed Toy', 'Candle', 'Sailboat', 'Laptop', 'Awning',
+        'Bed', 'Faucet', 'Tent', 'Horse', 'Mirror', 'Power outlet', 'Sink', 'Apple', 'Air Conditioner', 'Knife',
+        'Hockey Stick', 'Paddle', 'Pickup Truck', 'Fork', 'Traffic Sign', 'Balloon', 'Tripod', 'Dog', 'Spoon', 'Clock',
+        'Pot', 'Cow', 'Cake', 'Dinning Table', 'Sheep', 'Hanger', 'Blackboard/Whiteboard', 'Napkin', 'Other Fish',
+        'Orange/Tangerine', 'Toiletry', 'Keyboard', 'Tomato', 'Lantern', 'Machinery Vehicle', 'Fan',
+        'Green Vegetables', 'Banana', 'Baseball Glove', 'Airplane', 'Mouse', 'Train', 'Pumpkin', 'Soccer', 'Skiboard',
+        'Luggage', 'Nightstand', 'Tea pot', 'Telephone', 'Trolley', 'Head Phone', 'Sports Car', 'Stop Sign',
+        'Dessert', 'Scooter', 'Stroller', 'Crane', 'Remote', 'Refrigerator', 'Oven', 'Lemon', 'Duck', 'Baseball Bat',
+        'Surveillance Camera', 'Cat', 'Jug', 'Broccoli', 'Piano', 'Pizza', 'Elephant', 'Skateboard', 'Surfboard',
+        'Gun', 'Skating and Skiing shoes', 'Gas stove', 'Donut', 'Bow Tie', 'Carrot', 'Toilet', 'Kite', 'Strawberry',
+        'Other Balls', 'Shovel', 'Pepper', 'Computer Box', 'Toilet Paper', 'Cleaning Products', 'Chopsticks',
+        'Microwave', 'Pigeon', 'Baseball', 'Cutting/chopping Board', 'Coffee Table', 'Side Table', 'Scissors',
+        'Marker', 'Pie', 'Ladder', 'Snowboard', 'Cookies', 'Radiator', 'Fire Hydrant', 'Basketball', 'Zebra', 'Grape',
+        'Giraffe', 'Potato', 'Sausage', 'Tricycle', 'Violin', 'Egg', 'Fire Extinguisher', 'Candy', 'Fire Truck',
+        'Billiards', 'Converter', 'Bathtub', 'Wheelchair', 'Golf Club', 'Briefcase', 'Cucumber', 'Cigar/Cigarette',
+        'Paint Brush', 'Pear', 'Heavy Truck', 'Hamburger', 'Extractor', 'Extension Cord', 'Tong', 'Tennis Racket',
+        'Folder', 'American Football', 'earphone', 'Mask', 'Kettle', 'Tennis', 'Ship', 'Swing', 'Coffee Machine',
+        'Slide', 'Carriage', 'Onion', 'Green beans', 'Projector', 'Frisbee', 'Washing Machine/Drying Machine',
+        'Chicken', 'Printer', 'Watermelon', 'Saxophone', 'Tissue', 'Toothbrush', 'Ice cream', 'Hot-air balloon',
+        'Cello', 'French Fries', 'Scale', 'Trophy', 'Cabbage', 'Hot dog', 'Blender', 'Peach', 'Rice', 'Wallet/Purse',
+        'Volleyball', 'Deer', 'Goose', 'Tape', 'Tablet', 'Cosmetics', 'Trumpet', 'Pineapple', 'Golf Ball',
+        'Ambulance', 'Parking meter', 'Mango', 'Key', 'Hurdle', 'Fishing Rod', 'Medal', 'Flute', 'Brush', 'Penguin',
+        'Megaphone', 'Corn', 'Lettuce', 'Garlic', 'Swan', 'Helicopter', 'Green Onion', 'Sandwich', 'Nuts',
+        'Speed Limit Sign', 'Induction Cooker', 'Broom', 'Trombone', 'Plum', 'Rickshaw', 'Goldfish', 'Kiwi fruit',
+        'Router/modem', 'Poker Card', 'Toaster', 'Shrimp', 'Sushi', 'Cheese', 'Notepaper', 'Cherry', 'Pliers', 'CD',
+        'Pasta', 'Hammer', 'Cue', 'Avocado', 'Hamimelon', 'Flask', 'Mushroom', 'Screwdriver', 'Soap', 'Recorder',
+        'Bear', 'Eggplant', 'Board Eraser', 'Coconut', 'Tape Measure/Ruler', 'Pig', 'Showerhead', 'Globe', 'Chips',
+        'Steak', 'Crosswalk Sign', 'Stapler', 'Camel', 'Formula 1', 'Pomegranate', 'Dishwasher', 'Crab',
+        'Hoverboard', 'Meat ball', 'Rice Cooker', 'Tuba', 'Calculator', 'Papaya', 'Antelope', 'Parrot', 'Seal',
+        'Butterfly', 'Dumbbell', 'Donkey', 'Lion', 'Urinal', 'Dolphin', 'Electric Drill', 'Hair Dryer', 'Egg tart',
+        'Jellyfish', 'Treadmill', 'Lighter', 'Grapefruit', 'Game board', 'Mop', 'Radish', 'Baozi', 'Target', 'French',
+        'Spring Rolls', 'Monkey', 'Rabbit', 'Pencil Case', 'Yak', 'Red Cabbage', 'Binoculars', 'Asparagus', 'Barbell',
+        'Scallop', 'Noddles', 'Comb', 'Dumpling', 'Oyster', 'Table Tennis paddle', 'Cosmetics Brush/Eyeliner Pencil',
+        'Chainsaw', 'Eraser', 'Lobster', 'Durian', 'Okra', 'Lipstick', 'Cosmetics Mirror', 'Curling', 'Table Tennis']
+
+
+# Download script/URL (optional) ---------------------------------------------------------------------------------------
+download: |
+  from pycocotools.coco import COCO
+  from tqdm import tqdm
+
+  from utils.general import Path, download, np, xyxy2xywhn
+
+  # Make Directories
+  dir = Path(yaml['path'])  # dataset root dir
+  for p in 'images', 'labels':
+      (dir / p).mkdir(parents=True, exist_ok=True)
+      for q in 'train', 'val':
+          (dir / p / q).mkdir(parents=True, exist_ok=True)
+
+  # Train, Val Splits
+  for split, patches in [('train', 50 + 1), ('val', 43 + 1)]:
+      print(f"Processing {split} in {patches} patches ...")
+      images, labels = dir / 'images' / split, dir / 'labels' / split
+
+      # Download
+      url = f"https://dorc.ks3-cn-beijing.ksyun.com/data-set/2020Objects365%E6%95%B0%E6%8D%AE%E9%9B%86/{split}/"
+      if split == 'train':
+          download([f'{url}zhiyuan_objv2_{split}.tar.gz'], dir=dir, delete=False)  # annotations json
+          download([f'{url}patch{i}.tar.gz' for i in range(patches)], dir=images, curl=True, delete=False, threads=8)
+      elif split == 'val':
+          download([f'{url}zhiyuan_objv2_{split}.json'], dir=dir, delete=False)  # annotations json
+          download([f'{url}images/v1/patch{i}.tar.gz' for i in range(15 + 1)], dir=images, curl=True, delete=False, threads=8)
+          download([f'{url}images/v2/patch{i}.tar.gz' for i in range(16, patches)], dir=images, curl=True, delete=False, threads=8)
+
+      # Move
+      for f in tqdm(images.rglob('*.jpg'), desc=f'Moving {split} images'):
+          f.rename(images / f.name)  # move to /images/{split}
+
+      # Labels
+      coco = COCO(dir / f'zhiyuan_objv2_{split}.json')
+      names = [x["name"] for x in coco.loadCats(coco.getCatIds())]
+      for cid, cat in enumerate(names):
+          catIds = coco.getCatIds(catNms=[cat])
+          imgIds = coco.getImgIds(catIds=catIds)
+          for im in tqdm(coco.loadImgs(imgIds), desc=f'Class {cid + 1}/{len(names)} {cat}'):
+              width, height = im["width"], im["height"]
+              path = Path(im["file_name"])  # image filename
+              try:
+                  with open(labels / path.with_suffix('.txt').name, 'a') as file:
+                      annIds = coco.getAnnIds(imgIds=im["id"], catIds=catIds, iscrowd=None)
+                      for a in coco.loadAnns(annIds):
+                          x, y, w, h = a['bbox']  # bounding box in xywh (xy top-left corner)
+                          xyxy = np.array([x, y, x + w, y + h])[None]  # pixels(1,4)
+                          x, y, w, h = xyxy2xywhn(xyxy, w=width, h=height, clip=True)[0]  # normalized and clipped
+                          file.write(f"{cid} {x:.5f} {y:.5f} {w:.5f} {h:.5f}\n")
+              except Exception as e:
+                  print(e)

data/SKU-110K.yaml

@@ -0,0 +1,52 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+# SKU-110K retail items dataset https://github.com/eg4000/SKU110K_CVPR19 by Trax Retail
+# Example usage: python train.py --data SKU-110K.yaml
+# parent
+# ├── yolov5
+# └── datasets
+#     └── SKU-110K  ← downloads here
+
+
+# Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..]
+path: ../datasets/SKU-110K  # dataset root dir
+train: train.txt  # train images (relative to 'path')  8219 images
+val: val.txt  # val images (relative to 'path')  588 images
+test: test.txt  # test images (optional)  2936 images
+
+# Classes
+nc: 1  # number of classes
+names: ['object']  # class names
+
+
+# Download script/URL (optional) ---------------------------------------------------------------------------------------
+download: |
+  import shutil
+  from tqdm import tqdm
+  from utils.general import np, pd, Path, download, xyxy2xywh
+
+  # Download
+  dir = Path(yaml['path'])  # dataset root dir
+  parent = Path(dir.parent)  # download dir
+  urls = ['http://trax-geometry.s3.amazonaws.com/cvpr_challenge/SKU110K_fixed.tar.gz']
+  download(urls, dir=parent, delete=False)
+
+  # Rename directories
+  if dir.exists():
+      shutil.rmtree(dir)
+  (parent / 'SKU110K_fixed').rename(dir)  # rename dir
+  (dir / 'labels').mkdir(parents=True, exist_ok=True)  # create labels dir
+
+  # Convert labels
+  names = 'image', 'x1', 'y1', 'x2', 'y2', 'class', 'image_width', 'image_height'  # column names
+  for d in 'annotations_train.csv', 'annotations_val.csv', 'annotations_test.csv':
+      x = pd.read_csv(dir / 'annotations' / d, names=names).values  # annotations
+      images, unique_images = x[:, 0], np.unique(x[:, 0])
+      with open((dir / d).with_suffix('.txt').__str__().replace('annotations_', ''), 'w') as f:
+          f.writelines(f'./images/{s}\n' for s in unique_images)
+      for im in tqdm(unique_images, desc=f'Converting {dir / d}'):
+          cls = 0  # single-class dataset
+          with open((dir / 'labels' / im).with_suffix('.txt'), 'a') as f:
+              for r in x[images == im]:
+                  w, h = r[6], r[7]  # image width, height
+                  xywh = xyxy2xywh(np.array([[r[1] / w, r[2] / h, r[3] / w, r[4] / h]]))[0]  # instance
+                  f.write(f"{cls} {xywh[0]:.5f} {xywh[1]:.5f} {xywh[2]:.5f} {xywh[3]:.5f}\n")  # write label

data/VOC.yaml

@@ -0,0 +1,80 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+# PASCAL VOC dataset http://host.robots.ox.ac.uk/pascal/VOC by University of Oxford
+# Example usage: python train.py --data VOC.yaml
+# parent
+# ├── yolov5
+# └── datasets
+#     └── VOC  ← downloads here
+
+
+# Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..]
+path: /root/autodl-tmp/yolov5-6.1/datasets/VOC2007_wm
+train: # train images (relative to 'path')  16551 images
+  # - images/train2012
+  - train.txt
+  # - images/val2012
+  - val.txt
+val: # val images (relative to 'path')  4952 images
+  - val.txt
+test: # test images (optional)
+  - test.txt
+
+# Classes
+nc: 20  # number of classes
+names: ['aeroplane', 'bicycle', 'bird', 'boat', 'bottle', 'bus', 'car', 'cat', 'chair', 'cow', 'diningtable', 'dog',
+        'horse', 'motorbike', 'person', 'pottedplant', 'sheep', 'sofa', 'train', 'tvmonitor']  # class names
+
+
+# Download script/URL (optional) ---------------------------------------------------------------------------------------
+download: |
+  import xml.etree.ElementTree as ET
+
+  from tqdm import tqdm
+  from utils.general import download, Path
+
+
+  def convert_label(path, lb_path, year, image_id):
+      def convert_box(size, box):
+          dw, dh = 1. / size[0], 1. / size[1]
+          x, y, w, h = (box[0] + box[1]) / 2.0 - 1, (box[2] + box[3]) / 2.0 - 1, box[1] - box[0], box[3] - box[2]
+          return x * dw, y * dh, w * dw, h * dh
+
+      in_file = open(path / f'VOC{year}/Annotations/{image_id}.xml')
+      out_file = open(lb_path, 'w')
+      tree = ET.parse(in_file)
+      root = tree.getroot()
+      size = root.find('size')
+      w = int(size.find('width').text)
+      h = int(size.find('height').text)
+
+      for obj in root.iter('object'):
+          cls = obj.find('name').text
+          if cls in yaml['names'] and not int(obj.find('difficult').text) == 1:
+              xmlbox = obj.find('bndbox')
+              bb = convert_box((w, h), [float(xmlbox.find(x).text) for x in ('xmin', 'xmax', 'ymin', 'ymax')])
+              cls_id = yaml['names'].index(cls)  # class id
+              out_file.write(" ".join([str(a) for a in (cls_id, *bb)]) + '\n')
+
+
+  # Download
+  dir = Path(yaml['path'])  # dataset root dir
+  url = 'https://github.com/ultralytics/yolov5/releases/download/v1.0/'
+  urls = [url + 'VOCtrainval_06-Nov-2007.zip',  # 446MB, 5012 images
+          url + 'VOCtest_06-Nov-2007.zip',  # 438MB, 4953 images
+          url + 'VOCtrainval_11-May-2012.zip']  # 1.95GB, 17126 images
+  download(urls, dir=dir / 'images', delete=False)
+
+  # Convert
+  path = dir / f'images/VOCdevkit'
+  for year, image_set in ('2012', 'train'), ('2012', 'val'), ('2007', 'train'), ('2007', 'val'), ('2007', 'test'):
+      imgs_path = dir / 'images' / f'{image_set}{year}'
+      lbs_path = dir / 'labels' / f'{image_set}{year}'
+      imgs_path.mkdir(exist_ok=True, parents=True)
+      lbs_path.mkdir(exist_ok=True, parents=True)
+
+      image_ids = open(path / f'VOC{year}/ImageSets/Main/{image_set}.txt').read().strip().split()
+      for id in tqdm(image_ids, desc=f'{image_set}{year}'):
+          f = path / f'VOC{year}/JPEGImages/{id}.jpg'  # old img path
+          lb_path = (lbs_path / f.name).with_suffix('.txt')  # new label path
+          f.rename(imgs_path / f.name)  # move image
+          convert_label(path, lb_path, year, id)  # convert labels to YOLO format

data/VisDrone.yaml

@@ -0,0 +1,61 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+# VisDrone2019-DET dataset https://github.com/VisDrone/VisDrone-Dataset by Tianjin University
+# Example usage: python train.py --data VisDrone.yaml
+# parent
+# ├── yolov5
+# └── datasets
+#     └── VisDrone  ← downloads here
+
+
+# Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..]
+path: ../datasets/VisDrone  # dataset root dir
+train: VisDrone2019-DET-train/images  # train images (relative to 'path')  6471 images
+val: VisDrone2019-DET-val/images  # val images (relative to 'path')  548 images
+test: VisDrone2019-DET-test-dev/images  # test images (optional)  1610 images
+
+# Classes
+nc: 10  # number of classes
+names: ['pedestrian', 'people', 'bicycle', 'car', 'van', 'truck', 'tricycle', 'awning-tricycle', 'bus', 'motor']
+
+
+# Download script/URL (optional) ---------------------------------------------------------------------------------------
+download: |
+  from utils.general import download, os, Path
+
+  def visdrone2yolo(dir):
+      from PIL import Image
+      from tqdm import tqdm
+
+      def convert_box(size, box):
+          # Convert VisDrone box to YOLO xywh box
+          dw = 1. / size[0]
+          dh = 1. / size[1]
+          return (box[0] + box[2] / 2) * dw, (box[1] + box[3] / 2) * dh, box[2] * dw, box[3] * dh
+
+      (dir / 'labels').mkdir(parents=True, exist_ok=True)  # make labels directory
+      pbar = tqdm((dir / 'annotations').glob('*.txt'), desc=f'Converting {dir}')
+      for f in pbar:
+          img_size = Image.open((dir / 'images' / f.name).with_suffix('.jpg')).size
+          lines = []
+          with open(f, 'r') as file:  # read annotation.txt
+              for row in [x.split(',') for x in file.read().strip().splitlines()]:
+                  if row[4] == '0':  # VisDrone 'ignored regions' class 0
+                      continue
+                  cls = int(row[5]) - 1
+                  box = convert_box(img_size, tuple(map(int, row[:4])))
+                  lines.append(f"{cls} {' '.join(f'{x:.6f}' for x in box)}\n")
+                  with open(str(f).replace(os.sep + 'annotations' + os.sep, os.sep + 'labels' + os.sep), 'w') as fl:
+                      fl.writelines(lines)  # write label.txt
+
+
+  # Download
+  dir = Path(yaml['path'])  # dataset root dir
+  urls = ['https://github.com/ultralytics/yolov5/releases/download/v1.0/VisDrone2019-DET-train.zip',
+          'https://github.com/ultralytics/yolov5/releases/download/v1.0/VisDrone2019-DET-val.zip',
+          'https://github.com/ultralytics/yolov5/releases/download/v1.0/VisDrone2019-DET-test-dev.zip',
+          'https://github.com/ultralytics/yolov5/releases/download/v1.0/VisDrone2019-DET-test-challenge.zip']
+  download(urls, dir=dir)
+
+  # Convert
+  for d in 'VisDrone2019-DET-train', 'VisDrone2019-DET-val', 'VisDrone2019-DET-test-dev':
+      visdrone2yolo(dir / d)  # convert VisDrone annotations to YOLO labels

data/coco.yaml

@@ -0,0 +1,44 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+# COCO 2017 dataset http://cocodataset.org by Microsoft
+# Example usage: python train.py --data coco.yaml
+# parent
+# ├── yolov5
+# └── datasets
+#     └── coco  ← downloads here
+
+
+# Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..]
+path: ../datasets/coco  # dataset root dir
+train: train2017.txt  # train images (relative to 'path') 118287 images
+val: val2017.txt  # val images (relative to 'path') 5000 images
+test: test-dev2017.txt  # 20288 of 40670 images, submit to https://competitions.codalab.org/competitions/20794
+
+# Classes
+nc: 80  # number of classes
+names: ['person', 'bicycle', 'car', 'motorcycle', 'airplane', 'bus', 'train', 'truck', 'boat', 'traffic light',
+        'fire hydrant', 'stop sign', 'parking meter', 'bench', 'bird', 'cat', 'dog', 'horse', 'sheep', 'cow',
+        'elephant', 'bear', 'zebra', 'giraffe', 'backpack', 'umbrella', 'handbag', 'tie', 'suitcase', 'frisbee',
+        'skis', 'snowboard', 'sports ball', 'kite', 'baseball bat', 'baseball glove', 'skateboard', 'surfboard',
+        'tennis racket', 'bottle', 'wine glass', 'cup', 'fork', 'knife', 'spoon', 'bowl', 'banana', 'apple',
+        'sandwich', 'orange', 'broccoli', 'carrot', 'hot dog', 'pizza', 'donut', 'cake', 'chair', 'couch',
+        'potted plant', 'bed', 'dining table', 'toilet', 'tv', 'laptop', 'mouse', 'remote', 'keyboard', 'cell phone',
+        'microwave', 'oven', 'toaster', 'sink', 'refrigerator', 'book', 'clock', 'vase', 'scissors', 'teddy bear',
+        'hair drier', 'toothbrush']  # class names
+
+
+# Download script/URL (optional)
+download: |
+  from utils.general import download, Path
+
+  # Download labels
+  segments = False  # segment or box labels
+  dir = Path(yaml['path'])  # dataset root dir
+  url = 'https://github.com/ultralytics/yolov5/releases/download/v1.0/'
+  urls = [url + ('coco2017labels-segments.zip' if segments else 'coco2017labels.zip')]  # labels
+  download(urls, dir=dir.parent)
+
+  # Download data
+  urls = ['http://images.cocodataset.org/zips/train2017.zip',  # 19G, 118k images
+          'http://images.cocodataset.org/zips/val2017.zip',  # 1G, 5k images
+          'http://images.cocodataset.org/zips/test2017.zip']  # 7G, 41k images (optional)
+  download(urls, dir=dir / 'images', threads=3)

data/coco128.yaml

@@ -0,0 +1,30 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+# COCO128 dataset https://www.kaggle.com/ultralytics/coco128 (first 128 images from COCO train2017) by Ultralytics
+# Example usage: python train.py --data coco128.yaml
+# parent
+# ├── yolov5
+# └── datasets
+#     └── coco128  ← downloads here
+
+
+# Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..]
+path: ../datasets/coco128  # dataset root dir
+train: images/train2017  # train images (relative to 'path') 128 images
+val: images/train2017  # val images (relative to 'path') 128 images
+test:  # test images (optional)
+
+# Classes
+nc: 80  # number of classes
+names: ['person', 'bicycle', 'car', 'motorcycle', 'airplane', 'bus', 'train', 'truck', 'boat', 'traffic light',
+        'fire hydrant', 'stop sign', 'parking meter', 'bench', 'bird', 'cat', 'dog', 'horse', 'sheep', 'cow',
+        'elephant', 'bear', 'zebra', 'giraffe', 'backpack', 'umbrella', 'handbag', 'tie', 'suitcase', 'frisbee',
+        'skis', 'snowboard', 'sports ball', 'kite', 'baseball bat', 'baseball glove', 'skateboard', 'surfboard',
+        'tennis racket', 'bottle', 'wine glass', 'cup', 'fork', 'knife', 'spoon', 'bowl', 'banana', 'apple',
+        'sandwich', 'orange', 'broccoli', 'carrot', 'hot dog', 'pizza', 'donut', 'cake', 'chair', 'couch',
+        'potted plant', 'bed', 'dining table', 'toilet', 'tv', 'laptop', 'mouse', 'remote', 'keyboard', 'cell phone',
+        'microwave', 'oven', 'toaster', 'sink', 'refrigerator', 'book', 'clock', 'vase', 'scissors', 'teddy bear',
+        'hair drier', 'toothbrush']  # class names
+
+
+# Download script/URL (optional)
+download: https://ultralytics.com/assets/coco128.zip

data/hyps/hyp.Objects365.yaml

@@ -0,0 +1,34 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+# Hyperparameters for Objects365 training
+# python train.py --weights yolov5m.pt --data Objects365.yaml --evolve
+# See Hyperparameter Evolution tutorial for details https://github.com/ultralytics/yolov5#tutorials
+
+lr0: 0.00258
+lrf: 0.17
+momentum: 0.779
+weight_decay: 0.00058
+warmup_epochs: 1.33
+warmup_momentum: 0.86
+warmup_bias_lr: 0.0711
+box: 0.0539
+cls: 0.299
+cls_pw: 0.825
+obj: 0.632
+obj_pw: 1.0
+iou_t: 0.2
+anchor_t: 3.44
+anchors: 3.2
+fl_gamma: 0.0
+hsv_h: 0.0188
+hsv_s: 0.704
+hsv_v: 0.36
+degrees: 0.0
+translate: 0.0902
+scale: 0.491
+shear: 0.0
+perspective: 0.0
+flipud: 0.0
+fliplr: 0.5
+mosaic: 1.0
+mixup: 0.0
+copy_paste: 0.0

data/hyps/hyp.VOC.yaml

@@ -0,0 +1,40 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+# Hyperparameters for VOC training
+# python train.py --batch 128 --weights yolov5m6.pt --data VOC.yaml --epochs 50 --img 512 --hyp hyp.scratch-med.yaml --evolve
+# See Hyperparameter Evolution tutorial for details https://github.com/ultralytics/yolov5#tutorials
+
+# YOLOv5 Hyperparameter Evolution Results
+# Best generation: 319
+# Last generation: 434
+#    metrics/precision,       metrics/recall,      metrics/mAP_0.5, metrics/mAP_0.5:0.95,         val/box_loss,         val/obj_loss,         val/cls_loss
+#              0.86236,              0.86184,              0.91274,              0.72647,            0.0077056,            0.0042449,            0.0013846
+
+lr0: 0.0033
+lrf: 0.15184
+momentum: 0.74747
+weight_decay: 0.00025
+warmup_epochs: 3.4278
+warmup_momentum: 0.59032
+warmup_bias_lr: 0.18742
+box: 0.02
+cls: 0.21563
+cls_pw: 0.5
+obj: 0.50843
+obj_pw: 0.6729
+iou_t: 0.2
+anchor_t: 3.4172
+fl_gamma: 0.0
+hsv_h: 0.01032
+hsv_s: 0.5562
+hsv_v: 0.28255
+degrees: 0.0
+translate: 0.04575
+scale: 0.73711
+shear: 0.0
+perspective: 0.0
+flipud: 0.0
+fliplr: 0.5
+mosaic: 0.87158
+mixup: 0.04294
+copy_paste: 0.0
+anchors: 3.3556

data/hyps/hyp.scratch-high.yaml

@@ -0,0 +1,34 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+# Hyperparameters for high-augmentation COCO training from scratch
+# python train.py --batch 32 --cfg yolov5m6.yaml --weights '' --data coco.yaml --img 1280 --epochs 300
+# See tutorials for hyperparameter evolution https://github.com/ultralytics/yolov5#tutorials
+
+lr0: 0.01  # initial learning rate (SGD=1E-2, Adam=1E-3)
+lrf: 0.1  # final OneCycleLR learning rate (lr0 * lrf)
+momentum: 0.937  # SGD momentum/Adam beta1
+weight_decay: 0.0005  # optimizer weight decay 5e-4
+warmup_epochs: 3.0  # warmup epochs (fractions ok)
+warmup_momentum: 0.8  # warmup initial momentum
+warmup_bias_lr: 0.1  # warmup initial bias lr
+box: 0.05  # box loss gain
+cls: 0.3  # cls loss gain
+cls_pw: 1.0  # cls BCELoss positive_weight
+obj: 0.7  # obj loss gain (scale with pixels)
+obj_pw: 1.0  # obj BCELoss positive_weight
+iou_t: 0.20  # IoU training threshold
+anchor_t: 4.0  # anchor-multiple threshold
+# anchors: 3  # anchors per output layer (0 to ignore)
+fl_gamma: 0.0  # focal loss gamma (efficientDet default gamma=1.5)
+hsv_h: 0.015  # image HSV-Hue augmentation (fraction)
+hsv_s: 0.7  # image HSV-Saturation augmentation (fraction)
+hsv_v: 0.4  # image HSV-Value augmentation (fraction)
+degrees: 0.0  # image rotation (+/- deg)
+translate: 0.1  # image translation (+/- fraction)
+scale: 0.9  # image scale (+/- gain)
+shear: 0.0  # image shear (+/- deg)
+perspective: 0.0  # image perspective (+/- fraction), range 0-0.001
+flipud: 0.0  # image flip up-down (probability)
+fliplr: 0.5  # image flip left-right (probability)
+mosaic: 1.0  # image mosaic (probability)
+mixup: 0.1  # image mixup (probability)
+copy_paste: 0.1  # segment copy-paste (probability)

data/hyps/hyp.scratch-low.yaml

@@ -0,0 +1,34 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+# Hyperparameters for low-augmentation COCO training from scratch
+# python train.py --batch 64 --cfg yolov5n6.yaml --weights '' --data coco.yaml --img 640 --epochs 300 --linear
+# See tutorials for hyperparameter evolution https://github.com/ultralytics/yolov5#tutorials
+
+lr0: 0.01  # initial learning rate (SGD=1E-2, Adam=1E-3)
+lrf: 0.01  # final OneCycleLR learning rate (lr0 * lrf)
+momentum: 0.937  # SGD momentum/Adam beta1
+weight_decay: 0.0005  # optimizer weight decay 5e-4
+warmup_epochs: 3.0  # warmup epochs (fractions ok)
+warmup_momentum: 0.8  # warmup initial momentum
+warmup_bias_lr: 0.1  # warmup initial bias lr
+box: 0.05  # box loss gain
+cls: 0.5  # cls loss gain
+cls_pw: 1.0  # cls BCELoss positive_weight
+obj: 1.0  # obj loss gain (scale with pixels)
+obj_pw: 1.0  # obj BCELoss positive_weight
+iou_t: 0.20  # IoU training threshold
+anchor_t: 4.0  # anchor-multiple threshold
+# anchors: 3  # anchors per output layer (0 to ignore)
+fl_gamma: 0.0  # focal loss gamma (efficientDet default gamma=1.5)
+hsv_h: 0.015  # image HSV-Hue augmentation (fraction)
+hsv_s: 0.7  # image HSV-Saturation augmentation (fraction)
+hsv_v: 0.4  # image HSV-Value augmentation (fraction)
+degrees: 0.0  # image rotation (+/- deg)
+translate: 0.1  # image translation (+/- fraction)
+scale: 0.5  # image scale (+/- gain)
+shear: 0.0  # image shear (+/- deg)
+perspective: 0.0  # image perspective (+/- fraction), range 0-0.001
+flipud: 0.0  # image flip up-down (probability)
+fliplr: 0.5  # image flip left-right (probability)
+mosaic: 1.0  # image mosaic (probability)
+mixup: 0.0  # image mixup (probability)
+copy_paste: 0.0  # segment copy-paste (probability)

data/hyps/hyp.scratch-med.yaml

@@ -0,0 +1,34 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+# Hyperparameters for medium-augmentation COCO training from scratch
+# python train.py --batch 32 --cfg yolov5m6.yaml --weights '' --data coco.yaml --img 1280 --epochs 300
+# See tutorials for hyperparameter evolution https://github.com/ultralytics/yolov5#tutorials
+
+lr0: 0.01  # initial learning rate (SGD=1E-2, Adam=1E-3)
+lrf: 0.1  # final OneCycleLR learning rate (lr0 * lrf)
+momentum: 0.937  # SGD momentum/Adam beta1
+weight_decay: 0.0005  # optimizer weight decay 5e-4
+warmup_epochs: 3.0  # warmup epochs (fractions ok)
+warmup_momentum: 0.8  # warmup initial momentum
+warmup_bias_lr: 0.1  # warmup initial bias lr
+box: 0.05  # box loss gain
+cls: 0.3  # cls loss gain
+cls_pw: 1.0  # cls BCELoss positive_weight
+obj: 0.7  # obj loss gain (scale with pixels)
+obj_pw: 1.0  # obj BCELoss positive_weight
+iou_t: 0.20  # IoU training threshold
+anchor_t: 4.0  # anchor-multiple threshold
+# anchors: 3  # anchors per output layer (0 to ignore)
+fl_gamma: 0.0  # focal loss gamma (efficientDet default gamma=1.5)
+hsv_h: 0.015  # image HSV-Hue augmentation (fraction)
+hsv_s: 0.7  # image HSV-Saturation augmentation (fraction)
+hsv_v: 0.4  # image HSV-Value augmentation (fraction)
+degrees: 0.0  # image rotation (+/- deg)
+translate: 0.1  # image translation (+/- fraction)
+scale: 0.9  # image scale (+/- gain)
+shear: 0.0  # image shear (+/- deg)
+perspective: 0.0  # image perspective (+/- fraction), range 0-0.001
+flipud: 0.0  # image flip up-down (probability)
+fliplr: 0.5  # image flip left-right (probability)
+mosaic: 1.0  # image mosaic (probability)
+mixup: 0.1  # image mixup (probability)
+copy_paste: 0.0  # segment copy-paste (probability)

data/scripts/download_weights.sh

@@ -0,0 +1,20 @@
+#!/bin/bash
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+# Download latest models from https://github.com/ultralytics/yolov5/releases
+# Example usage: bash path/to/download_weights.sh
+# parent
+# └── yolov5
+#     ├── yolov5s.pt  ← downloads here
+#     ├── yolov5m.pt
+#     └── ...
+
+python - <<EOF
+from utils.downloads import attempt_download
+
+models = ['n', 's', 'm', 'l', 'x']
+models.extend([x + '6' for x in models])  # add P6 models
+
+for x in models:
+    attempt_download(f'yolov5{x}.pt')
+
+EOF

data/scripts/get_coco.sh

@@ -0,0 +1,27 @@
+#!/bin/bash
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+# Download COCO 2017 dataset http://cocodataset.org
+# Example usage: bash data/scripts/get_coco.sh
+# parent
+# ├── yolov5
+# └── datasets
+#     └── coco  ← downloads here
+
+# Download/unzip labels
+d='../datasets' # unzip directory
+url=https://github.com/ultralytics/yolov5/releases/download/v1.0/
+f='coco2017labels.zip' # or 'coco2017labels-segments.zip', 68 MB
+echo 'Downloading' $url$f ' ...'
+curl -L $url$f -o $f && unzip -q $f -d $d && rm $f &
+
+# Download/unzip images
+d='../datasets/coco/images' # unzip directory
+url=http://images.cocodataset.org/zips/
+f1='train2017.zip' # 19G, 118k images
+f2='val2017.zip'   # 1G, 5k images
+f3='test2017.zip'  # 7G, 41k images (optional)
+for f in $f1 $f2; do
+  echo 'Downloading' $url$f '...'
+  curl -L $url$f -o $f && unzip -q $f -d $d && rm $f &
+done
+wait # finish background tasks

data/scripts/get_coco128.sh

@@ -0,0 +1,17 @@
+#!/bin/bash
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+# Download COCO128 dataset https://www.kaggle.com/ultralytics/coco128 (first 128 images from COCO train2017)
+# Example usage: bash data/scripts/get_coco128.sh
+# parent
+# ├── yolov5
+# └── datasets
+#     └── coco128  ← downloads here
+
+# Download/unzip images and labels
+d='../datasets' # unzip directory
+url=https://github.com/ultralytics/yolov5/releases/download/v1.0/
+f='coco128.zip' # or 'coco128-segments.zip', 68 MB
+echo 'Downloading' $url$f ' ...'
+curl -L $url$f -o $f && unzip -q $f -d $d && rm $f &
+
+wait # finish background tasks

data/voc_ball.yaml

@@ -0,0 +1,76 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+# PASCAL VOC dataset http://host.robots.ox.ac.uk/pascal/VOC by University of Oxford
+# Example usage: python train.py --data VOC.yaml
+# parent
+# ├── yolov5
+# └── datasets
+#     └── VOC  ← downloads here
+
+
+# Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..]
+path: ./
+train: # train images (relative to 'path')  16551 images
+  - VOCdevkit/images/train/ 
+val: # val images (relative to 'path')  4952 images
+  - VOCdevkit/images/val/
+test: # test images (optional)
+
+
+# Classes
+nc: 1  # number of classes
+names: ['ball']  # class names
+
+
+# Download script/URL (optional) ---------------------------------------------------------------------------------------
+download: |
+  import xml.etree.ElementTree as ET
+
+  from tqdm import tqdm
+  from utils.general import download, Path
+
+
+  def convert_label(path, lb_path, year, image_id):
+      def convert_box(size, box):
+          dw, dh = 1. / size[0], 1. / size[1]
+          x, y, w, h = (box[0] + box[1]) / 2.0 - 1, (box[2] + box[3]) / 2.0 - 1, box[1] - box[0], box[3] - box[2]
+          return x * dw, y * dh, w * dw, h * dh
+
+      in_file = open(path / f'VOC{year}/Annotations/{image_id}.xml')
+      out_file = open(lb_path, 'w')
+      tree = ET.parse(in_file)
+      root = tree.getroot()
+      size = root.find('size')
+      w = int(size.find('width').text)
+      h = int(size.find('height').text)
+
+      for obj in root.iter('object'):
+          cls = obj.find('name').text
+          if cls in yaml['names'] and not int(obj.find('difficult').text) == 1:
+              xmlbox = obj.find('bndbox')
+              bb = convert_box((w, h), [float(xmlbox.find(x).text) for x in ('xmin', 'xmax', 'ymin', 'ymax')])
+              cls_id = yaml['names'].index(cls)  # class id
+              out_file.write(" ".join([str(a) for a in (cls_id, *bb)]) + '\n')
+
+
+  # Download
+  dir = Path(yaml['path'])  # dataset root dir
+  url = 'https://github.com/ultralytics/yolov5/releases/download/v1.0/'
+  urls = [url + 'VOCtrainval_06-Nov-2007.zip',  # 446MB, 5012 images
+          url + 'VOCtest_06-Nov-2007.zip',  # 438MB, 4953 images
+          url + 'VOCtrainval_11-May-2012.zip']  # 1.95GB, 17126 images
+  download(urls, dir=dir / 'images', delete=False)
+
+  # Convert
+  path = dir / f'images/VOCdevkit'
+  for year, image_set in ('2012', 'train'), ('2012', 'val'), ('2007', 'train'), ('2007', 'val'), ('2007', 'test'):
+      imgs_path = dir / 'images' / f'{image_set}{year}'
+      lbs_path = dir / 'labels' / f'{image_set}{year}'
+      imgs_path.mkdir(exist_ok=True, parents=True)
+      lbs_path.mkdir(exist_ok=True, parents=True)
+
+      image_ids = open(path / f'VOC{year}/ImageSets/Main/{image_set}.txt').read().strip().split()
+      for id in tqdm(image_ids, desc=f'{image_set}{year}'):
+          f = path / f'VOC{year}/JPEGImages/{id}.jpg'  # old img path
+          lb_path = (lbs_path / f.name).with_suffix('.txt')  # new label path
+          f.rename(imgs_path / f.name)  # move image
+          convert_label(path, lb_path, year, id)  # convert labels to YOLO format

data/xView.yaml

@@ -0,0 +1,102 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+# DIUx xView 2018 Challenge https://challenge.xviewdataset.org by U.S. National Geospatial-Intelligence Agency (NGA)
+# --------  DOWNLOAD DATA MANUALLY and jar xf val_images.zip to 'datasets/xView' before running train command!  --------
+# Example usage: python train.py --data xView.yaml
+# parent
+# ├── yolov5
+# └── datasets
+#     └── xView  ← downloads here
+
+
+# Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..]
+path: ../datasets/xView  # dataset root dir
+train: images/autosplit_train.txt  # train images (relative to 'path') 90% of 847 train images
+val: images/autosplit_val.txt  # train images (relative to 'path') 10% of 847 train images
+
+# Classes
+nc: 60  # number of classes
+names: ['Fixed-wing Aircraft', 'Small Aircraft', 'Cargo Plane', 'Helicopter', 'Passenger Vehicle', 'Small Car', 'Bus',
+        'Pickup Truck', 'Utility Truck', 'Truck', 'Cargo Truck', 'Truck w/Box', 'Truck Tractor', 'Trailer',
+        'Truck w/Flatbed', 'Truck w/Liquid', 'Crane Truck', 'Railway Vehicle', 'Passenger Car', 'Cargo Car',
+        'Flat Car', 'Tank car', 'Locomotive', 'Maritime Vessel', 'Motorboat', 'Sailboat', 'Tugboat', 'Barge',
+        'Fishing Vessel', 'Ferry', 'Yacht', 'Container Ship', 'Oil Tanker', 'Engineering Vehicle', 'Tower crane',
+        'Container Crane', 'Reach Stacker', 'Straddle Carrier', 'Mobile Crane', 'Dump Truck', 'Haul Truck',
+        'Scraper/Tractor', 'Front loader/Bulldozer', 'Excavator', 'Cement Mixer', 'Ground Grader', 'Hut/Tent', 'Shed',
+        'Building', 'Aircraft Hangar', 'Damaged Building', 'Facility', 'Construction Site', 'Vehicle Lot', 'Helipad',
+        'Storage Tank', 'Shipping container lot', 'Shipping Container', 'Pylon', 'Tower']  # class names
+
+
+# Download script/URL (optional) ---------------------------------------------------------------------------------------
+download: |
+  import json
+  import os
+  from pathlib import Path
+
+  import numpy as np
+  from PIL import Image
+  from tqdm import tqdm
+
+  from utils.datasets import autosplit
+  from utils.general import download, xyxy2xywhn
+
+
+  def convert_labels(fname=Path('xView/xView_train.geojson')):
+      # Convert xView geoJSON labels to YOLO format
+      path = fname.parent
+      with open(fname) as f:
+          print(f'Loading {fname}...')
+          data = json.load(f)
+
+      # Make dirs
+      labels = Path(path / 'labels' / 'train')
+      os.system(f'rm -rf {labels}')
+      labels.mkdir(parents=True, exist_ok=True)
+
+      # xView classes 11-94 to 0-59
+      xview_class2index = [-1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 0, 1, 2, -1, 3, -1, 4, 5, 6, 7, 8, -1, 9, 10, 11,
+                           12, 13, 14, 15, -1, -1, 16, 17, 18, 19, 20, 21, 22, -1, 23, 24, 25, -1, 26, 27, -1, 28, -1,
+                           29, 30, 31, 32, 33, 34, 35, 36, 37, -1, 38, 39, 40, 41, 42, 43, 44, 45, -1, -1, -1, -1, 46,
+                           47, 48, 49, -1, 50, 51, -1, 52, -1, -1, -1, 53, 54, -1, 55, -1, -1, 56, -1, 57, -1, 58, 59]
+
+      shapes = {}
+      for feature in tqdm(data['features'], desc=f'Converting {fname}'):
+          p = feature['properties']
+          if p['bounds_imcoords']:
+              id = p['image_id']
+              file = path / 'train_images' / id
+              if file.exists():  # 1395.tif missing
+                  try:
+                      box = np.array([int(num) for num in p['bounds_imcoords'].split(",")])
+                      assert box.shape[0] == 4, f'incorrect box shape {box.shape[0]}'
+                      cls = p['type_id']
+                      cls = xview_class2index[int(cls)]  # xView class to 0-60
+                      assert 59 >= cls >= 0, f'incorrect class index {cls}'
+
+                      # Write YOLO label
+                      if id not in shapes:
+                          shapes[id] = Image.open(file).size
+                      box = xyxy2xywhn(box[None].astype(np.float), w=shapes[id][0], h=shapes[id][1], clip=True)
+                      with open((labels / id).with_suffix('.txt'), 'a') as f:
+                          f.write(f"{cls} {' '.join(f'{x:.6f}' for x in box[0])}\n")  # write label.txt
+                  except Exception as e:
+                      print(f'WARNING: skipping one label for {file}: {e}')
+
+
+  # Download manually from https://challenge.xviewdataset.org
+  dir = Path(yaml['path'])  # dataset root dir
+  # urls = ['https://d307kc0mrhucc3.cloudfront.net/train_labels.zip',  # train labels
+  #         'https://d307kc0mrhucc3.cloudfront.net/train_images.zip',  # 15G, 847 train images
+  #         'https://d307kc0mrhucc3.cloudfront.net/val_images.zip']  # 5G, 282 val images (no labels)
+  # download(urls, dir=dir, delete=False)
+
+  # Convert labels
+  convert_labels(dir / 'xView_train.geojson')
+
+  # Move images
+  images = Path(dir / 'images')
+  images.mkdir(parents=True, exist_ok=True)
+  Path(dir / 'train_images').rename(dir / 'images' / 'train')
+  Path(dir / 'val_images').rename(dir / 'images' / 'val')
+
+  # Split
+  autosplit(dir / 'images' / 'train')

detect.py

@@ -0,0 +1,257 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+"""
+Run inference on images, videos, directories, streams, etc.
+
+Usage - sources:
+    $ python path/to/detect.py --weights yolov5s.pt --source 0              # webcam
+                                                             img.jpg        # image
+                                                             vid.mp4        # video
+                                                             path/          # directory
+                                                             path/*.jpg     # glob
+                                                             'https://youtu.be/Zgi9g1ksQHc'  # YouTube
+                                                             'rtsp://example.com/media.mp4'  # RTSP, RTMP, HTTP stream
+
+Usage - formats:
+    $ python path/to/detect.py --weights yolov5s.pt                 # PyTorch
+                                         yolov5s.torchscript        # TorchScript
+                                         yolov5s.onnx               # ONNX Runtime or OpenCV DNN with --dnn
+                                         yolov5s.xml                # OpenVINO
+                                         yolov5s.engine             # TensorRT
+                                         yolov5s.mlmodel            # CoreML (MacOS-only)
+                                         yolov5s_saved_model        # TensorFlow SavedModel
+                                         yolov5s.pb                 # TensorFlow GraphDef
+                                         yolov5s.tflite             # TensorFlow Lite
+                                         yolov5s_edgetpu.tflite     # TensorFlow Edge TPU
+"""
+
+import argparse
+import os
+import sys
+from pathlib import Path
+
+import cv2
+import torch
+import torch.backends.cudnn as cudnn
+
+FILE = Path(__file__).resolve()
+ROOT = FILE.parents[0]  # YOLOv5 root directory
+if str(ROOT) not in sys.path:
+    sys.path.append(str(ROOT))  # add ROOT to PATH
+ROOT = Path(os.path.relpath(ROOT, Path.cwd()))  # relative
+
+from models.common import DetectMultiBackend
+from utils.datasets import IMG_FORMATS, VID_FORMATS, LoadImages, LoadStreams
+from utils.general import (LOGGER, check_file, check_img_size, check_imshow, check_requirements, colorstr,
+                           increment_path, non_max_suppression, print_args, scale_coords, strip_optimizer, xyxy2xywh)
+from utils.plots import Annotator, colors, save_one_box
+from utils.torch_utils import select_device, time_sync
+
+
+@torch.no_grad()
+def run(weights=ROOT / 'yolov5s.pt',  # model.pt path(s)
+        source=ROOT / 'data/images',  # file/dir/URL/glob, 0 for webcam
+        data=ROOT / 'data/coco128.yaml',  # dataset.yaml path
+        imgsz=(640, 640),  # inference size (height, width)
+        conf_thres=0.25,  # confidence threshold
+        iou_thres=0.45,  # NMS IOU threshold
+        max_det=1000,  # maximum detections per image
+        device='',  # cuda device, i.e. 0 or 0,1,2,3 or cpu
+        view_img=False,  # show results
+        save_txt=False,  # save results to *.txt
+        save_conf=False,  # save confidences in --save-txt labels
+        save_crop=False,  # save cropped prediction boxes
+        nosave=False,  # do not save images/videos
+        classes=None,  # filter by class: --class 0, or --class 0 2 3
+        agnostic_nms=False,  # class-agnostic NMS
+        augment=False,  # augmented inference
+        visualize=False,  # visualize features
+        update=False,  # update all models
+        project=ROOT / 'runs/detect',  # save results to project/name
+        name='exp',  # save results to project/name
+        exist_ok=False,  # existing project/name ok, do not increment
+        line_thickness=3,  # bounding box thickness (pixels)
+        hide_labels=False,  # hide labels
+        hide_conf=False,  # hide confidences
+        half=False,  # use FP16 half-precision inference
+        dnn=False,  # use OpenCV DNN for ONNX inference
+        ):
+    source = str(source)
+    save_img = not nosave and not source.endswith('.txt')  # save inference images
+    is_file = Path(source).suffix[1:] in (IMG_FORMATS + VID_FORMATS)
+    is_url = source.lower().startswith(('rtsp://', 'rtmp://', 'http://', 'https://'))
+    webcam = source.isnumeric() or source.endswith('.txt') or (is_url and not is_file)
+    if is_url and is_file:
+        source = check_file(source)  # download
+
+    # Directories
+    save_dir = increment_path(Path(project) / name, exist_ok=exist_ok)  # increment run
+    (save_dir / 'labels' if save_txt else save_dir).mkdir(parents=True, exist_ok=True)  # make dir
+
+    # Load model
+    device = select_device(device)
+    model = DetectMultiBackend(weights, device=device, dnn=dnn, data=data)
+    stride, names, pt, jit, onnx, engine = model.stride, model.names, model.pt, model.jit, model.onnx, model.engine
+    imgsz = check_img_size(imgsz, s=stride)  # check image size
+
+    # Half
+    half &= (pt or jit or onnx or engine) and device.type != 'cpu'  # FP16 supported on limited backends with CUDA
+    if pt or jit:
+        model.model.half() if half else model.model.float()
+
+    # Dataloader
+    if webcam:
+        view_img = check_imshow()
+        cudnn.benchmark = True  # set True to speed up constant image size inference
+        dataset = LoadStreams(source, img_size=imgsz, stride=stride, auto=pt)
+        bs = len(dataset)  # batch_size
+    else:
+        dataset = LoadImages(source, img_size=imgsz, stride=stride, auto=pt)
+        bs = 1  # batch_size
+    vid_path, vid_writer = [None] * bs, [None] * bs
+
+    # Run inference
+    model.warmup(imgsz=(1 if pt else bs, 3, *imgsz), half=half)  # warmup
+    dt, seen = [0.0, 0.0, 0.0], 0
+    for path, im, im0s, vid_cap, s in dataset:
+        t1 = time_sync()
+        im = torch.from_numpy(im).to(device)
+        im = im.half() if half else im.float()  # uint8 to fp16/32
+        im /= 255  # 0 - 255 to 0.0 - 1.0
+        if len(im.shape) == 3:
+            im = im[None]  # expand for batch dim
+        t2 = time_sync()
+        dt[0] += t2 - t1
+
+        # Inference
+        visualize = increment_path(save_dir / Path(path).stem, mkdir=True) if visualize else False
+        pred = model(im, augment=augment, visualize=visualize)
+        t3 = time_sync()
+        dt[1] += t3 - t2
+
+        # NMS
+        pred = non_max_suppression(pred, conf_thres, iou_thres, classes, agnostic_nms, max_det=max_det)
+        dt[2] += time_sync() - t3
+
+        # Second-stage classifier (optional)
+        # pred = utils.general.apply_classifier(pred, classifier_model, im, im0s)
+
+        # Process predictions
+        for i, det in enumerate(pred):  # per image
+            seen += 1
+            if webcam:  # batch_size >= 1
+                p, im0, frame = path[i], im0s[i].copy(), dataset.count
+                s += f'{i}: '
+            else:
+                p, im0, frame = path, im0s.copy(), getattr(dataset, 'frame', 0)
+
+            p = Path(p)  # to Path
+            save_path = str(save_dir / p.name)  # im.jpg
+            txt_path = str(save_dir / 'labels' / p.stem) + ('' if dataset.mode == 'image' else f'_{frame}')  # im.txt
+            s += '%gx%g ' % im.shape[2:]  # print string
+            gn = torch.tensor(im0.shape)[[1, 0, 1, 0]]  # normalization gain whwh
+            imc = im0.copy() if save_crop else im0  # for save_crop
+            annotator = Annotator(im0, line_width=line_thickness, example=str(names))
+            if len(det):
+                # Rescale boxes from img_size to im0 size
+                det[:, :4] = scale_coords(im.shape[2:], det[:, :4], im0.shape).round()
+
+                # Print results
+                for c in det[:, -1].unique():
+                    n = (det[:, -1] == c).sum()  # detections per class
+                    s += f"{n} {names[int(c)]}{'s' * (n > 1)}, "  # add to string
+
+                # Write results
+                for *xyxy, conf, cls in reversed(det):
+                    if save_txt:  # Write to file
+                        xywh = (xyxy2xywh(torch.tensor(xyxy).view(1, 4)) / gn).view(-1).tolist()  # normalized xywh
+                        line = (cls, *xywh, conf) if save_conf else (cls, *xywh)  # label format
+                        with open(txt_path + '.txt', 'a') as f:
+                            f.write(('%g ' * len(line)).rstrip() % line + '\n')
+
+                    if save_img or save_crop or view_img:  # Add bbox to image
+                        c = int(cls)  # integer class
+                        label = None if hide_labels else (names[c] if hide_conf else f'{names[c]} {conf:.2f}')
+                        annotator.box_label(xyxy, label, color=colors(c, True))
+                        if save_crop:
+                            save_one_box(xyxy, imc, file=save_dir / 'crops' / names[c] / f'{p.stem}.jpg', BGR=True)
+
+            # Stream results
+            im0 = annotator.result()
+            if view_img:
+                cv2.imshow(str(p), im0)
+                cv2.waitKey(1)  # 1 millisecond
+
+            # Save results (image with detections)
+            if save_img:
+                if dataset.mode == 'image':
+                    cv2.imwrite(save_path, im0)
+                else:  # 'video' or 'stream'
+                    if vid_path[i] != save_path:  # new video
+                        vid_path[i] = save_path
+                        if isinstance(vid_writer[i], cv2.VideoWriter):
+                            vid_writer[i].release()  # release previous video writer
+                        if vid_cap:  # video
+                            fps = vid_cap.get(cv2.CAP_PROP_FPS)
+                            w = int(vid_cap.get(cv2.CAP_PROP_FRAME_WIDTH))
+                            h = int(vid_cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
+                        else:  # stream
+                            fps, w, h = 30, im0.shape[1], im0.shape[0]
+                        save_path = str(Path(save_path).with_suffix('.mp4'))  # force *.mp4 suffix on results videos
+                        vid_writer[i] = cv2.VideoWriter(save_path, cv2.VideoWriter_fourcc(*'mp4v'), fps, (w, h))
+                    vid_writer[i].write(im0)
+
+        # Print time (inference-only)
+        LOGGER.info(f'{s}Done. ({t3 - t2:.3f}s)')
+
+    # Print results
+    t = tuple(x / seen * 1E3 for x in dt)  # speeds per image
+    LOGGER.info(f'Speed: %.1fms pre-process, %.1fms inference, %.1fms NMS per image at shape {(1, 3, *imgsz)}' % t)
+    if save_txt or save_img:
+        s = f"\n{len(list(save_dir.glob('labels/*.txt')))} labels saved to {save_dir / 'labels'}" if save_txt else ''
+        LOGGER.info(f"Results saved to {colorstr('bold', save_dir)}{s}")
+    if update:
+        strip_optimizer(weights)  # update model (to fix SourceChangeWarning)
+
+
+def parse_opt():
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--weights', nargs='+', type=str, default=ROOT / 'VOC2007_wm/train/exp5/weights/best.pt', help='model path(s)')
+    parser.add_argument('--source', type=str, default=ROOT / 'datasets/VOC2007_wm/images', help='file/dir/URL/glob, 0 for webcam')
+    parser.add_argument('--data', type=str, default=ROOT / 'data/VOC.yaml', help='(optional) dataset.yaml path')
+    parser.add_argument('--imgsz', '--img', '--img-size', nargs='+', type=int, default=[640], help='inference size h,w')
+    parser.add_argument('--conf-thres', type=float, default=0.25, help='confidence threshold')
+    parser.add_argument('--iou-thres', type=float, default=0.45, help='NMS IoU threshold')
+    parser.add_argument('--max-det', type=int, default=1000, help='maximum detections per image')
+    parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')
+    parser.add_argument('--view-img', action='store_true', help='show results')
+    parser.add_argument('--save-txt', action='store_true', help='save results to *.txt')
+    parser.add_argument('--save-conf', action='store_true', help='save confidences in --save-txt labels')
+    parser.add_argument('--save-crop', action='store_true', help='save cropped prediction boxes')
+    parser.add_argument('--nosave', action='store_true', help='do not save images/videos')
+    parser.add_argument('--classes', nargs='+', type=int, help='filter by class: --classes 0, or --classes 0 2 3')
+    parser.add_argument('--agnostic-nms', action='store_true', help='class-agnostic NMS')
+    parser.add_argument('--augment', action='store_true', help='augmented inference')
+    parser.add_argument('--visualize', action='store_true', help='visualize features')
+    parser.add_argument('--update', action='store_true', help='update all models')
+    parser.add_argument('--project', default=ROOT / 'VOC2007_wm/detect', help='save results to project/name')
+    parser.add_argument('--name', default='exp', help='save results to project/name')
+    parser.add_argument('--exist-ok', action='store_true', help='existing project/name ok, do not increment')
+    parser.add_argument('--line-thickness', default=3, type=int, help='bounding box thickness (pixels)')
+    parser.add_argument('--hide-labels', default=False, action='store_true', help='hide labels')
+    parser.add_argument('--hide-conf', default=False, action='store_true', help='hide confidences')
+    parser.add_argument('--half', action='store_true', help='use FP16 half-precision inference')
+    parser.add_argument('--dnn', action='store_true', help='use OpenCV DNN for ONNX inference')
+    opt = parser.parse_args()
+    opt.imgsz *= 2 if len(opt.imgsz) == 1 else 1  # expand
+    print_args(FILE.stem, opt)
+    return opt
+
+
+def main(opt):
+    check_requirements(exclude=('tensorboard', 'thop'))
+    run(**vars(opt))
+
+
+if __name__ == "__main__":
+    opt = parse_opt()
+    main(opt)

detect_pruned.py

@@ -0,0 +1,258 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+"""
+Run inference on images, videos, directories, streams, etc.
+
+Usage - sources:
+    $ python path/to/detect.py --weights yolov5s.pt --source 0  # webcam
+                                                             img.jpg  # image
+                                                             vid.mp4  # video
+                                                             path/  # directory
+                                                             path/*.jpg  # glob
+                                                             'https://youtu.be/Zgi9g1ksQHc'  # YouTube
+                                                             'rtsp://example.com/media.mp4'  # RTSP, RTMP, HTTP stream
+
+Usage - formats:
+    $ python path/to/detect.py --weights yolov5s.pt                 # PyTorch
+                                         yolov5s.torchscript        # TorchScript
+                                         yolov5s.onnx               # ONNX Runtime or OpenCV DNN with --dnn
+                                         yolov5s.xml                # OpenVINO
+                                         yolov5s.engine             # TensorRT
+                                         yolov5s.mlmodel            # CoreML (MacOS-only)
+                                         yolov5s_saved_model        # TensorFlow SavedModel
+                                         yolov5s.pb                 # TensorFlow GraphDef
+                                         yolov5s.tflite             # TensorFlow Lite
+                                         yolov5s_edgetpu.tflite     # TensorFlow Edge TPU
+"""
+
+import argparse
+import os
+import sys
+from pathlib import Path
+
+import cv2
+import torch
+import torch.backends.cudnn as cudnn
+
+FILE = Path(__file__).resolve()
+ROOT = FILE.parents[0]  # YOLOv5 root directory
+if str(ROOT) not in sys.path:
+    sys.path.append(str(ROOT))  # add ROOT to PATH
+ROOT = Path(os.path.relpath(ROOT, Path.cwd()))  # relative
+
+from models.common import DetectPrunedMultiBackend
+from utils.datasets import IMG_FORMATS, VID_FORMATS, LoadImages, LoadStreams
+from utils.general import (LOGGER, check_file, check_img_size, check_imshow, check_requirements, colorstr,
+                           increment_path, non_max_suppression, print_args, scale_coords, strip_optimizer, xyxy2xywh)
+from utils.plots import Annotator, colors, save_one_box
+from utils.torch_utils import select_device, time_sync
+
+
+@torch.no_grad()
+def run(weights=ROOT / 'yolov5s.pt',  # model.pt path(s)
+        source=ROOT / 'data/images',  # file/dir/URL/glob, 0 for webcam
+        data=ROOT / 'data/coco128.yaml',  # dataset.yaml path
+        imgsz=(640, 640),  # inference size (height, width)
+        conf_thres=0.25,  # confidence threshold
+        iou_thres=0.45,  # NMS IOU threshold
+        max_det=1000,  # maximum detections per image
+        device='',  # cuda device, i.e. 0 or 0,1,2,3 or cpu
+        view_img=False,  # show results
+        save_txt=False,  # save results to *.txt
+        save_conf=False,  # save confidences in --save-txt labels
+        save_crop=False,  # save cropped prediction boxes
+        nosave=False,  # do not save images/videos
+        classes=None,  # filter by class: --class 0, or --class 0 2 3
+        agnostic_nms=False,  # class-agnostic NMS
+        augment=False,  # augmented inference
+        visualize=False,  # visualize features
+        update=False,  # update all models
+        project=ROOT / 'runs/detect',  # save results to project/name
+        name='exp',  # save results to project/name
+        exist_ok=False,  # existing project/name ok, do not increment
+        line_thickness=3,  # bounding box thickness (pixels)
+        hide_labels=False,  # hide labels
+        hide_conf=False,  # hide confidences
+        half=False,  # use FP16 half-precision inference
+        dnn=False,  # use OpenCV DNN for ONNX inference
+        ):
+    source = str(source)
+    save_img = not nosave and not source.endswith('.txt')  # save inference images
+    is_file = Path(source).suffix[1:] in (IMG_FORMATS + VID_FORMATS)
+    is_url = source.lower().startswith(('rtsp://', 'rtmp://', 'http://', 'https://'))
+    webcam = source.isnumeric() or source.endswith('.txt') or (is_url and not is_file)
+    if is_url and is_file:
+        source = check_file(source)  # download
+
+    # Directories
+    save_dir = increment_path(Path(project) / name, exist_ok=exist_ok)  # increment run
+    (save_dir / 'labels' if save_txt else save_dir).mkdir(parents=True, exist_ok=True)  # make dir
+
+    # Load model
+    device = select_device(device)
+    print("weights:",weights)
+    model = DetectPrunedMultiBackend(weights, device=device, dnn=dnn)
+    stride, names, pt, jit, onnx, engine = model.stride, model.names, model.pt, model.jit, model.onnx, model.engine
+    imgsz = check_img_size(imgsz, s=stride)  # check image size
+
+    # Half
+    half &= (pt or jit or onnx or engine) and device.type != 'cpu'  # FP16 supported on limited backends with CUDA
+    if pt or jit:
+        model.model.half() if half else model.model.float()
+
+    # Dataloader
+    if webcam:
+        view_img = check_imshow()
+        cudnn.benchmark = True  # set True to speed up constant image size inference
+        dataset = LoadStreams(source, img_size=imgsz, stride=stride, auto=pt)
+        bs = len(dataset)  # batch_size
+    else:
+        dataset = LoadImages(source, img_size=imgsz, stride=stride, auto=pt)
+        bs = 1  # batch_size
+    vid_path, vid_writer = [None] * bs, [None] * bs
+
+    # Run inference
+    model.warmup(imgsz=(1 if pt else bs, 3, *imgsz), half=half)  # warmup
+    dt, seen = [0.0, 0.0, 0.0], 0
+    for path, im, im0s, vid_cap, s in dataset:
+        t1 = time_sync()
+        im = torch.from_numpy(im).to(device)
+        im = im.half() if half else im.float()  # uint8 to fp16/32
+        im /= 255  # 0 - 255 to 0.0 - 1.0
+        if len(im.shape) == 3:
+            im = im[None]  # expand for batch dim
+        t2 = time_sync()
+        dt[0] += t2 - t1
+
+        # Inference
+        visualize = increment_path(save_dir / Path(path).stem, mkdir=True) if visualize else False
+        pred = model(im, augment=augment, visualize=visualize)
+        t3 = time_sync()
+        dt[1] += t3 - t2
+
+        # NMS
+        pred = non_max_suppression(pred, conf_thres, iou_thres, classes, agnostic_nms, max_det=max_det)
+        dt[2] += time_sync() - t3
+
+        # Second-stage classifier (optional)
+        # pred = utils.general.apply_classifier(pred, classifier_model, im, im0s)
+
+        # Process predictions
+        for i, det in enumerate(pred):  # per image
+            seen += 1
+            if webcam:  # batch_size >= 1
+                p, im0, frame = path[i], im0s[i].copy(), dataset.count
+                s += f'{i}: '
+            else:
+                p, im0, frame = path, im0s.copy(), getattr(dataset, 'frame', 0)
+
+            p = Path(p)  # to Path
+            save_path = str(save_dir / p.name)  # im.jpg
+            txt_path = str(save_dir / 'labels' / p.stem) + ('' if dataset.mode == 'image' else f'_{frame}')  # im.txt
+            s += '%gx%g ' % im.shape[2:]  # print string
+            gn = torch.tensor(im0.shape)[[1, 0, 1, 0]]  # normalization gain whwh
+            imc = im0.copy() if save_crop else im0  # for save_crop
+            annotator = Annotator(im0, line_width=line_thickness, example=str(names))
+            if len(det):
+                # Rescale boxes from img_size to im0 size
+                det[:, :4] = scale_coords(im.shape[2:], det[:, :4], im0.shape).round()
+
+                # Print results
+                for c in det[:, -1].unique():
+                    n = (det[:, -1] == c).sum()  # detections per class
+                    s += f"{n} {names[int(c)]}{'s' * (n > 1)}, "  # add to string
+
+                # Write results
+                for *xyxy, conf, cls in reversed(det):
+                    if save_txt:  # Write to file
+                        xywh = (xyxy2xywh(torch.tensor(xyxy).view(1, 4)) / gn).view(-1).tolist()  # normalized xywh
+                        line = (cls, *xywh, conf) if save_conf else (cls, *xywh)  # label format
+                        with open(txt_path + '.txt', 'a') as f:
+                            f.write(('%g ' * len(line)).rstrip() % line + '\n')
+
+                    if save_img or save_crop or view_img:  # Add bbox to image
+                        c = int(cls)  # integer class
+                        label = None if hide_labels else (names[c] if hide_conf else f'{names[c]} {conf:.2f}')
+                        annotator.box_label(xyxy, label, color=colors(c, True))
+                        if save_crop:
+                            save_one_box(xyxy, imc, file=save_dir / 'crops' / names[c] / f'{p.stem}.jpg', BGR=True)
+
+            # Stream results
+            im0 = annotator.result()
+            if view_img:
+                cv2.imshow(str(p), im0)
+                cv2.waitKey(1)  # 1 millisecond
+
+            # Save results (image with detections)
+            if save_img:
+                if dataset.mode == 'image':
+                    cv2.imwrite(save_path, im0)
+                else:  # 'video' or 'stream'
+                    if vid_path[i] != save_path:  # new video
+                        vid_path[i] = save_path
+                        if isinstance(vid_writer[i], cv2.VideoWriter):
+                            vid_writer[i].release()  # release previous video writer
+                        if vid_cap:  # video
+                            fps = vid_cap.get(cv2.CAP_PROP_FPS)
+                            w = int(vid_cap.get(cv2.CAP_PROP_FRAME_WIDTH))
+                            h = int(vid_cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
+                        else:  # stream
+                            fps, w, h = 30, im0.shape[1], im0.shape[0]
+                        save_path = str(Path(save_path).with_suffix('.mp4'))  # force *.mp4 suffix on results videos
+                        vid_writer[i] = cv2.VideoWriter(save_path, cv2.VideoWriter_fourcc(*'mp4v'), fps, (w, h))
+                    vid_writer[i].write(im0)
+
+        # Print time (inference-only)
+        LOGGER.info(f'{s}Done. ({t3 - t2:.3f}s)')
+
+    # Print results
+    t = tuple(x / seen * 1E3 for x in dt)  # speeds per image
+    LOGGER.info(f'Speed: %.1fms pre-process, %.1fms inference, %.1fms NMS per image at shape {(1, 3, *imgsz)}' % t)
+    if save_txt or save_img:
+        s = f"\n{len(list(save_dir.glob('labels/*.txt')))} labels saved to {save_dir / 'labels'}" if save_txt else ''
+        LOGGER.info(f"Results saved to {colorstr('bold', save_dir)}{s}")
+    if update:
+        strip_optimizer(weights)  # update model (to fix SourceChangeWarning)
+
+
+def parse_opt():
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--weights', nargs='+', type=str, default=ROOT / 'VOC2007_wm/finetune_pruned/exp/weights/best.pt', help='model path(s)')
+    parser.add_argument('--source', type=str, default=ROOT / 'datasets/VOC2007_wm/images', help='file/dir/URL/glob, 0 for webcam')
+    parser.add_argument('--data', type=str, default=ROOT / 'data/VOC.yaml', help='(optional) dataset.yaml path')
+    parser.add_argument('--imgsz', '--img', '--img-size', nargs='+', type=int, default=[640], help='inference size h,w')
+    parser.add_argument('--conf-thres', type=float, default=0.25, help='confidence threshold')
+    parser.add_argument('--iou-thres', type=float, default=0.45, help='NMS IoU threshold')
+    parser.add_argument('--max-det', type=int, default=1000, help='maximum detections per image')
+    parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')
+    parser.add_argument('--view-img', action='store_true', help='show results')
+    parser.add_argument('--save-txt', action='store_true', help='save results to *.txt')
+    parser.add_argument('--save-conf', action='store_true', help='save confidences in --save-txt labels')
+    parser.add_argument('--save-crop', action='store_true', help='save cropped prediction boxes')
+    parser.add_argument('--nosave', action='store_true', help='do not save images/videos')
+    parser.add_argument('--classes', nargs='+', type=int, help='filter by class: --classes 0, or --classes 0 2 3')
+    parser.add_argument('--agnostic-nms', action='store_true', help='class-agnostic NMS')
+    parser.add_argument('--augment', action='store_true', help='augmented inference')
+    parser.add_argument('--visualize', action='store_true', help='visualize features')
+    parser.add_argument('--update', action='store_true', help='update all models')
+    parser.add_argument('--project', default=ROOT / 'VOC2007_wm/finetune_pruned/detect', help='save results to project/name')
+    parser.add_argument('--name', default='exp', help='save results to project/name')
+    parser.add_argument('--exist-ok', action='store_true', help='existing project/name ok, do not increment')
+    parser.add_argument('--line-thickness', default=3, type=int, help='bounding box thickness (pixels)')
+    parser.add_argument('--hide-labels', default=False, action='store_true', help='hide labels')
+    parser.add_argument('--hide-conf', default=False, action='store_true', help='hide confidences')
+    parser.add_argument('--half', action='store_true', help='use FP16 half-precision inference')
+    parser.add_argument('--dnn', action='store_true', help='use OpenCV DNN for ONNX inference')
+    opt = parser.parse_args()
+    opt.imgsz *= 2 if len(opt.imgsz) == 1 else 1  # expand
+    print_args(FILE.stem, opt)
+    return opt
+
+
+def main(opt):
+    check_requirements(exclude=('tensorboard', 'thop'))
+    run(**vars(opt))
+
+
+if __name__ == "__main__":
+    opt = parse_opt()
+    main(opt)

export.py

@@ -0,0 +1,559 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+"""
+Export a YOLOv5 PyTorch model to other formats. TensorFlow exports authored by https://github.com/zldrobit
+
+Format                      | `export.py --include`         | Model
+---                         | ---                           | ---
+PyTorch                     | -                             | yolov5s.pt
+TorchScript                 | `torchscript`                 | yolov5s.torchscript
+ONNX                        | `onnx`                        | yolov5s.onnx
+OpenVINO                    | `openvino`                    | yolov5s_openvino_model/
+TensorRT                    | `engine`                      | yolov5s.engine
+CoreML                      | `coreml`                      | yolov5s.mlmodel
+TensorFlow SavedModel       | `saved_model`                 | yolov5s_saved_model/
+TensorFlow GraphDef         | `pb`                          | yolov5s.pb
+TensorFlow Lite             | `tflite`                      | yolov5s.tflite
+TensorFlow Edge TPU         | `edgetpu`                     | yolov5s_edgetpu.tflite
+TensorFlow.js               | `tfjs`                        | yolov5s_web_model/
+
+Requirements:
+    $ pip install -r requirements.txt coremltools onnx onnx-simplifier onnxruntime openvino-dev tensorflow-cpu  # CPU
+    $ pip install -r requirements.txt coremltools onnx onnx-simplifier onnxruntime-gpu openvino-dev tensorflow  # GPU
+
+Usage:
+    $ python path/to/export.py --weights yolov5s.pt --include torchscript onnx openvino engine coreml tflite ...
+
+Inference:
+    $ python path/to/detect.py --weights yolov5s.pt                 # PyTorch
+                                         yolov5s.torchscript        # TorchScript
+                                         yolov5s.onnx               # ONNX Runtime or OpenCV DNN with --dnn
+                                         yolov5s.xml                # OpenVINO
+                                         yolov5s.engine             # TensorRT
+                                         yolov5s.mlmodel            # CoreML (MacOS-only)
+                                         yolov5s_saved_model        # TensorFlow SavedModel
+                                         yolov5s.pb                 # TensorFlow GraphDef
+                                         yolov5s.tflite             # TensorFlow Lite
+                                         yolov5s_edgetpu.tflite     # TensorFlow Edge TPU
+
+TensorFlow.js:
+    $ cd .. && git clone https://github.com/zldrobit/tfjs-yolov5-example.git && cd tfjs-yolov5-example
+    $ npm install
+    $ ln -s ../../yolov5/yolov5s_web_model public/yolov5s_web_model
+    $ npm start
+"""
+
+import argparse
+import json
+import os
+import platform
+import subprocess
+import sys
+import time
+import warnings
+from pathlib import Path
+
+import pandas as pd
+import torch
+import torch.nn as nn
+from torch.utils.mobile_optimizer import optimize_for_mobile
+
+FILE = Path(__file__).resolve()
+ROOT = FILE.parents[0]  # YOLOv5 root directory
+if str(ROOT) not in sys.path:
+    sys.path.append(str(ROOT))  # add ROOT to PATH
+ROOT = Path(os.path.relpath(ROOT, Path.cwd()))  # relative
+
+from models.common import Conv
+from models.experimental import attempt_load
+from models.yolo import Detect
+from utils.activations import SiLU
+from utils.datasets import LoadImages
+from utils.general import (LOGGER, check_dataset, check_img_size, check_requirements, check_version, colorstr,
+                           file_size, print_args, url2file)
+from utils.torch_utils import select_device
+
+
+def export_formats():
+    # YOLOv5 export formats
+    x = [['PyTorch', '-', '.pt'],
+         ['TorchScript', 'torchscript', '.torchscript'],
+         ['ONNX', 'onnx', '.onnx'],
+         ['OpenVINO', 'openvino', '_openvino_model'],
+         ['TensorRT', 'engine', '.engine'],
+         ['CoreML', 'coreml', '.mlmodel'],
+         ['TensorFlow SavedModel', 'saved_model', '_saved_model'],
+         ['TensorFlow GraphDef', 'pb', '.pb'],
+         ['TensorFlow Lite', 'tflite', '.tflite'],
+         ['TensorFlow Edge TPU', 'edgetpu', '_edgetpu.tflite'],
+         ['TensorFlow.js', 'tfjs', '_web_model']]
+    return pd.DataFrame(x, columns=['Format', 'Argument', 'Suffix'])
+
+
+def export_torchscript(model, im, file, optimize, prefix=colorstr('TorchScript:')):
+    # YOLOv5 TorchScript model export
+    try:
+        LOGGER.info(f'\n{prefix} starting export with torch {torch.__version__}...')
+        f = file.with_suffix('.torchscript')
+
+        ts = torch.jit.trace(model, im, strict=False)
+        d = {"shape": im.shape, "stride": int(max(model.stride)), "names": model.names}
+        extra_files = {'config.txt': json.dumps(d)}  # torch._C.ExtraFilesMap()
+        if optimize:  # https://pytorch.org/tutorials/recipes/mobile_interpreter.html
+            optimize_for_mobile(ts)._save_for_lite_interpreter(str(f), _extra_files=extra_files)
+        else:
+            ts.save(str(f), _extra_files=extra_files)
+
+        LOGGER.info(f'{prefix} export success, saved as {f} ({file_size(f):.1f} MB)')
+        return f
+    except Exception as e:
+        LOGGER.info(f'{prefix} export failure: {e}')
+
+
+def export_onnx(model, im, file, opset, train, dynamic, simplify, prefix=colorstr('ONNX:')):
+    # YOLOv5 ONNX export
+    try:
+        check_requirements(('onnx',))
+        import onnx
+
+        LOGGER.info(f'\n{prefix} starting export with onnx {onnx.__version__}...')
+        f = file.with_suffix('.onnx')
+
+        torch.onnx.export(model, im, f, verbose=False, opset_version=opset,
+                          training=torch.onnx.TrainingMode.TRAINING if train else torch.onnx.TrainingMode.EVAL,
+                          do_constant_folding=not train,
+                          input_names=['images'],
+                          output_names=['output'],
+                          dynamic_axes={'images': {0: 'batch', 2: 'height', 3: 'width'},  # shape(1,3,640,640)
+                                        'output': {0: 'batch', 1: 'anchors'}  # shape(1,25200,85)
+                                        } if dynamic else None)
+
+        # Checks
+        model_onnx = onnx.load(f)  # load onnx model
+        onnx.checker.check_model(model_onnx)  # check onnx model
+        # LOGGER.info(onnx.helper.printable_graph(model_onnx.graph))  # print
+
+        # Simplify
+        if simplify:
+            try:
+                check_requirements(('onnx-simplifier',))
+                import onnxsim
+
+                LOGGER.info(f'{prefix} simplifying with onnx-simplifier {onnxsim.__version__}...')
+                model_onnx, check = onnxsim.simplify(
+                    model_onnx,
+                    dynamic_input_shape=dynamic,
+                    input_shapes={'images': list(im.shape)} if dynamic else None)
+                assert check, 'assert check failed'
+                onnx.save(model_onnx, f)
+            except Exception as e:
+                LOGGER.info(f'{prefix} simplifier failure: {e}')
+        LOGGER.info(f'{prefix} export success, saved as {f} ({file_size(f):.1f} MB)')
+        return f
+    except Exception as e:
+        LOGGER.info(f'{prefix} export failure: {e}')
+
+
+def export_openvino(model, im, file, prefix=colorstr('OpenVINO:')):
+    # YOLOv5 OpenVINO export
+    try:
+        check_requirements(('openvino-dev',))  # requires openvino-dev: https://pypi.org/project/openvino-dev/
+        import openvino.inference_engine as ie
+
+        LOGGER.info(f'\n{prefix} starting export with openvino {ie.__version__}...')
+        f = str(file).replace('.pt', '_openvino_model' + os.sep)
+
+        cmd = f"mo --input_model {file.with_suffix('.onnx')} --output_dir {f}"
+        subprocess.check_output(cmd, shell=True)
+
+        LOGGER.info(f'{prefix} export success, saved as {f} ({file_size(f):.1f} MB)')
+        return f
+    except Exception as e:
+        LOGGER.info(f'\n{prefix} export failure: {e}')
+
+
+def export_coreml(model, im, file, prefix=colorstr('CoreML:')):
+    # YOLOv5 CoreML export
+    try:
+        check_requirements(('coremltools',))
+        import coremltools as ct
+
+        LOGGER.info(f'\n{prefix} starting export with coremltools {ct.__version__}...')
+        f = file.with_suffix('.mlmodel')
+
+        ts = torch.jit.trace(model, im, strict=False)  # TorchScript model
+        ct_model = ct.convert(ts, inputs=[ct.ImageType('image', shape=im.shape, scale=1 / 255, bias=[0, 0, 0])])
+        ct_model.save(f)
+
+        LOGGER.info(f'{prefix} export success, saved as {f} ({file_size(f):.1f} MB)')
+        return ct_model, f
+    except Exception as e:
+        LOGGER.info(f'\n{prefix} export failure: {e}')
+        return None, None
+
+
+def export_engine(model, im, file, train, half, simplify, workspace=4, verbose=False, prefix=colorstr('TensorRT:')):
+    # YOLOv5 TensorRT export https://developer.nvidia.com/tensorrt
+    try:
+        check_requirements(('tensorrt',))
+        import tensorrt as trt
+
+        if trt.__version__[0] == '7':  # TensorRT 7 handling https://github.com/ultralytics/yolov5/issues/6012
+            grid = model.model[-1].anchor_grid
+            model.model[-1].anchor_grid = [a[..., :1, :1, :] for a in grid]
+            export_onnx(model, im, file, 12, train, False, simplify)  # opset 12
+            model.model[-1].anchor_grid = grid
+        else:  # TensorRT >= 8
+            check_version(trt.__version__, '8.0.0', hard=True)  # require tensorrt>=8.0.0
+            export_onnx(model, im, file, 13, train, False, simplify)  # opset 13
+        onnx = file.with_suffix('.onnx')
+
+        LOGGER.info(f'\n{prefix} starting export with TensorRT {trt.__version__}...')
+        assert im.device.type != 'cpu', 'export running on CPU but must be on GPU, i.e. `python export.py --device 0`'
+        assert onnx.exists(), f'failed to export ONNX file: {onnx}'
+        f = file.with_suffix('.engine')  # TensorRT engine file
+        logger = trt.Logger(trt.Logger.INFO)
+        if verbose:
+            logger.min_severity = trt.Logger.Severity.VERBOSE
+
+        builder = trt.Builder(logger)
+        config = builder.create_builder_config()
+        config.max_workspace_size = workspace * 1 << 30
+
+        flag = (1 << int(trt.NetworkDefinitionCreationFlag.EXPLICIT_BATCH))
+        network = builder.create_network(flag)
+        parser = trt.OnnxParser(network, logger)
+        if not parser.parse_from_file(str(onnx)):
+            raise RuntimeError(f'failed to load ONNX file: {onnx}')
+
+        inputs = [network.get_input(i) for i in range(network.num_inputs)]
+        outputs = [network.get_output(i) for i in range(network.num_outputs)]
+        LOGGER.info(f'{prefix} Network Description:')
+        for inp in inputs:
+            LOGGER.info(f'{prefix}\tinput "{inp.name}" with shape {inp.shape} and dtype {inp.dtype}')
+        for out in outputs:
+            LOGGER.info(f'{prefix}\toutput "{out.name}" with shape {out.shape} and dtype {out.dtype}')
+
+        half &= builder.platform_has_fast_fp16
+        LOGGER.info(f'{prefix} building FP{16 if half else 32} engine in {f}')
+        if half:
+            config.set_flag(trt.BuilderFlag.FP16)
+        with builder.build_engine(network, config) as engine, open(f, 'wb') as t:
+            t.write(engine.serialize())
+        LOGGER.info(f'{prefix} export success, saved as {f} ({file_size(f):.1f} MB)')
+        return f
+    except Exception as e:
+        LOGGER.info(f'\n{prefix} export failure: {e}')
+
+
+def export_saved_model(model, im, file, dynamic,
+                       tf_nms=False, agnostic_nms=False, topk_per_class=100, topk_all=100, iou_thres=0.45,
+                       conf_thres=0.25, keras=False, prefix=colorstr('TensorFlow SavedModel:')):
+    # YOLOv5 TensorFlow SavedModel export
+    try:
+        import tensorflow as tf
+        from tensorflow.python.framework.convert_to_constants import convert_variables_to_constants_v2
+
+        from models.tf import TFDetect, TFModel
+
+        LOGGER.info(f'\n{prefix} starting export with tensorflow {tf.__version__}...')
+        f = str(file).replace('.pt', '_saved_model')
+        batch_size, ch, *imgsz = list(im.shape)  # BCHW
+
+        tf_model = TFModel(cfg=model.yaml, model=model, nc=model.nc, imgsz=imgsz)
+        im = tf.zeros((batch_size, *imgsz, 3))  # BHWC order for TensorFlow
+        _ = tf_model.predict(im, tf_nms, agnostic_nms, topk_per_class, topk_all, iou_thres, conf_thres)
+        inputs = tf.keras.Input(shape=(*imgsz, 3), batch_size=None if dynamic else batch_size)
+        outputs = tf_model.predict(inputs, tf_nms, agnostic_nms, topk_per_class, topk_all, iou_thres, conf_thres)
+        keras_model = tf.keras.Model(inputs=inputs, outputs=outputs)
+        keras_model.trainable = False
+        keras_model.summary()
+        if keras:
+            keras_model.save(f, save_format='tf')
+        else:
+            m = tf.function(lambda x: keras_model(x))  # full model
+            spec = tf.TensorSpec(keras_model.inputs[0].shape, keras_model.inputs[0].dtype)
+            m = m.get_concrete_function(spec)
+            frozen_func = convert_variables_to_constants_v2(m)
+            tfm = tf.Module()
+            tfm.__call__ = tf.function(lambda x: frozen_func(x), [spec])
+            tfm.__call__(im)
+            tf.saved_model.save(
+                tfm,
+                f,
+                options=tf.saved_model.SaveOptions(experimental_custom_gradients=False) if
+                check_version(tf.__version__, '2.6') else tf.saved_model.SaveOptions())
+        LOGGER.info(f'{prefix} export success, saved as {f} ({file_size(f):.1f} MB)')
+        return keras_model, f
+    except Exception as e:
+        LOGGER.info(f'\n{prefix} export failure: {e}')
+        return None, None
+
+
+def export_pb(keras_model, im, file, prefix=colorstr('TensorFlow GraphDef:')):
+    # YOLOv5 TensorFlow GraphDef *.pb export https://github.com/leimao/Frozen_Graph_TensorFlow
+    try:
+        import tensorflow as tf
+        from tensorflow.python.framework.convert_to_constants import convert_variables_to_constants_v2
+
+        LOGGER.info(f'\n{prefix} starting export with tensorflow {tf.__version__}...')
+        f = file.with_suffix('.pb')
+
+        m = tf.function(lambda x: keras_model(x))  # full model
+        m = m.get_concrete_function(tf.TensorSpec(keras_model.inputs[0].shape, keras_model.inputs[0].dtype))
+        frozen_func = convert_variables_to_constants_v2(m)
+        frozen_func.graph.as_graph_def()
+        tf.io.write_graph(graph_or_graph_def=frozen_func.graph, logdir=str(f.parent), name=f.name, as_text=False)
+
+        LOGGER.info(f'{prefix} export success, saved as {f} ({file_size(f):.1f} MB)')
+        return f
+    except Exception as e:
+        LOGGER.info(f'\n{prefix} export failure: {e}')
+
+
+def export_tflite(keras_model, im, file, int8, data, ncalib, prefix=colorstr('TensorFlow Lite:')):
+    # YOLOv5 TensorFlow Lite export
+    try:
+        import tensorflow as tf
+
+        LOGGER.info(f'\n{prefix} starting export with tensorflow {tf.__version__}...')
+        batch_size, ch, *imgsz = list(im.shape)  # BCHW
+        f = str(file).replace('.pt', '-fp16.tflite')
+
+        converter = tf.lite.TFLiteConverter.from_keras_model(keras_model)
+        converter.target_spec.supported_ops = [tf.lite.OpsSet.TFLITE_BUILTINS]
+        converter.target_spec.supported_types = [tf.float16]
+        converter.optimizations = [tf.lite.Optimize.DEFAULT]
+        if int8:
+            from models.tf import representative_dataset_gen
+            dataset = LoadImages(check_dataset(data)['train'], img_size=imgsz, auto=False)  # representative data
+            converter.representative_dataset = lambda: representative_dataset_gen(dataset, ncalib)
+            converter.target_spec.supported_ops = [tf.lite.OpsSet.TFLITE_BUILTINS_INT8]
+            converter.target_spec.supported_types = []
+            converter.inference_input_type = tf.uint8  # or tf.int8
+            converter.inference_output_type = tf.uint8  # or tf.int8
+            converter.experimental_new_quantizer = False
+            f = str(file).replace('.pt', '-int8.tflite')
+
+        tflite_model = converter.convert()
+        open(f, "wb").write(tflite_model)
+        LOGGER.info(f'{prefix} export success, saved as {f} ({file_size(f):.1f} MB)')
+        return f
+    except Exception as e:
+        LOGGER.info(f'\n{prefix} export failure: {e}')
+
+
+def export_edgetpu(keras_model, im, file, prefix=colorstr('Edge TPU:')):
+    # YOLOv5 Edge TPU export https://coral.ai/docs/edgetpu/models-intro/
+    try:
+        cmd = 'edgetpu_compiler --version'
+        help_url = 'https://coral.ai/docs/edgetpu/compiler/'
+        assert platform.system() == 'Linux', f'export only supported on Linux. See {help_url}'
+        if subprocess.run(cmd + ' >/dev/null', shell=True).returncode != 0:
+            LOGGER.info(f'\n{prefix} export requires Edge TPU compiler. Attempting install from {help_url}')
+            sudo = subprocess.run('sudo --version >/dev/null', shell=True).returncode == 0  # sudo installed on system
+            for c in ['curl https://packages.cloud.google.com/apt/doc/apt-key.gpg | sudo apt-key add -',
+                      'echo "deb https://packages.cloud.google.com/apt coral-edgetpu-stable main" | sudo tee /etc/apt/sources.list.d/coral-edgetpu.list',
+                      'sudo apt-get update',
+                      'sudo apt-get install edgetpu-compiler']:
+                subprocess.run(c if sudo else c.replace('sudo ', ''), shell=True, check=True)
+        ver = subprocess.run(cmd, shell=True, capture_output=True, check=True).stdout.decode().split()[-1]
+
+        LOGGER.info(f'\n{prefix} starting export with Edge TPU compiler {ver}...')
+        f = str(file).replace('.pt', '-int8_edgetpu.tflite')  # Edge TPU model
+        f_tfl = str(file).replace('.pt', '-int8.tflite')  # TFLite model
+
+        cmd = f"edgetpu_compiler -s {f_tfl}"
+        subprocess.run(cmd, shell=True, check=True)
+
+        LOGGER.info(f'{prefix} export success, saved as {f} ({file_size(f):.1f} MB)')
+        return f
+    except Exception as e:
+        LOGGER.info(f'\n{prefix} export failure: {e}')
+
+
+def export_tfjs(keras_model, im, file, prefix=colorstr('TensorFlow.js:')):
+    # YOLOv5 TensorFlow.js export
+    try:
+        check_requirements(('tensorflowjs',))
+        import re
+
+        import tensorflowjs as tfjs
+
+        LOGGER.info(f'\n{prefix} starting export with tensorflowjs {tfjs.__version__}...')
+        f = str(file).replace('.pt', '_web_model')  # js dir
+        f_pb = file.with_suffix('.pb')  # *.pb path
+        f_json = f + '/model.json'  # *.json path
+
+        cmd = f'tensorflowjs_converter --input_format=tf_frozen_model ' \
+              f'--output_node_names="Identity,Identity_1,Identity_2,Identity_3" {f_pb} {f}'
+        subprocess.run(cmd, shell=True)
+
+        json = open(f_json).read()
+        with open(f_json, 'w') as j:  # sort JSON Identity_* in ascending order
+            subst = re.sub(
+                r'{"outputs": {"Identity.?.?": {"name": "Identity.?.?"}, '
+                r'"Identity.?.?": {"name": "Identity.?.?"}, '
+                r'"Identity.?.?": {"name": "Identity.?.?"}, '
+                r'"Identity.?.?": {"name": "Identity.?.?"}}}',
+                r'{"outputs": {"Identity": {"name": "Identity"}, '
+                r'"Identity_1": {"name": "Identity_1"}, '
+                r'"Identity_2": {"name": "Identity_2"}, '
+                r'"Identity_3": {"name": "Identity_3"}}}',
+                json)
+            j.write(subst)
+
+        LOGGER.info(f'{prefix} export success, saved as {f} ({file_size(f):.1f} MB)')
+        return f
+    except Exception as e:
+        LOGGER.info(f'\n{prefix} export failure: {e}')
+
+
+@torch.no_grad()
+def run(data=ROOT / 'data/coco128.yaml',  # 'dataset.yaml path'
+        weights=ROOT / 'yolov5s.pt',  # weights path
+        imgsz=(640, 640),  # image (height, width)
+        batch_size=1,  # batch size
+        device='cpu',  # cuda device, i.e. 0 or 0,1,2,3 or cpu
+        include=('torchscript', 'onnx'),  # include formats
+        half=False,  # FP16 half-precision export
+        inplace=False,  # set YOLOv5 Detect() inplace=True
+        train=False,  # model.train() mode
+        optimize=False,  # TorchScript: optimize for mobile
+        int8=False,  # CoreML/TF INT8 quantization
+        dynamic=False,  # ONNX/TF: dynamic axes
+        simplify=False,  # ONNX: simplify model
+        opset=12,  # ONNX: opset version
+        verbose=False,  # TensorRT: verbose log
+        workspace=4,  # TensorRT: workspace size (GB)
+        nms=False,  # TF: add NMS to model
+        agnostic_nms=False,  # TF: add agnostic NMS to model
+        topk_per_class=100,  # TF.js NMS: topk per class to keep
+        topk_all=100,  # TF.js NMS: topk for all classes to keep
+        iou_thres=0.45,  # TF.js NMS: IoU threshold
+        conf_thres=0.25  # TF.js NMS: confidence threshold
+        ):
+    t = time.time()
+    include = [x.lower() for x in include]  # to lowercase
+    formats = tuple(export_formats()['Argument'][1:])  # --include arguments
+    flags = [x in include for x in formats]
+    assert sum(flags) == len(include), f'ERROR: Invalid --include {include}, valid --include arguments are {formats}'
+    jit, onnx, xml, engine, coreml, saved_model, pb, tflite, edgetpu, tfjs = flags  # export booleans
+    file = Path(url2file(weights) if str(weights).startswith(('http:/', 'https:/')) else weights)  # PyTorch weights
+
+    # Load PyTorch model
+    device = select_device(device)
+    assert not (device.type == 'cpu' and half), '--half only compatible with GPU export, i.e. use --device 0'
+    model = attempt_load(weights, map_location=device, inplace=True, fuse=True)  # load FP32 model
+    nc, names = model.nc, model.names  # number of classes, class names
+
+    # Checks
+    imgsz *= 2 if len(imgsz) == 1 else 1  # expand
+    opset = 12 if ('openvino' in include) else opset  # OpenVINO requires opset <= 12
+    assert nc == len(names), f'Model class count {nc} != len(names) {len(names)}'
+
+    # Input
+    gs = int(max(model.stride))  # grid size (max stride)
+    imgsz = [check_img_size(x, gs) for x in imgsz]  # verify img_size are gs-multiples
+    im = torch.zeros(batch_size, 3, *imgsz).to(device)  # image size(1,3,320,192) BCHW iDetection
+
+    # Update model
+    if half:
+        im, model = im.half(), model.half()  # to FP16
+    model.train() if train else model.eval()  # training mode = no Detect() layer grid construction
+    for k, m in model.named_modules():
+        if isinstance(m, Conv):  # assign export-friendly activations
+            if isinstance(m.act, nn.SiLU):
+                m.act = SiLU()
+        elif isinstance(m, Detect):
+            m.inplace = inplace
+            m.onnx_dynamic = dynamic
+            if hasattr(m, 'forward_export'):
+                m.forward = m.forward_export  # assign custom forward (optional)
+
+    for _ in range(2):
+        y = model(im)  # dry runs
+    shape = tuple(y[0].shape)  # model output shape
+    LOGGER.info(f"\n{colorstr('PyTorch:')} starting from {file} with output shape {shape} ({file_size(file):.1f} MB)")
+
+    # Exports
+    f = [''] * 10  # exported filenames
+    warnings.filterwarnings(action='ignore', category=torch.jit.TracerWarning)  # suppress TracerWarning
+    if jit:
+        f[0] = export_torchscript(model, im, file, optimize)
+    if engine:  # TensorRT required before ONNX
+        f[1] = export_engine(model, im, file, train, half, simplify, workspace, verbose)
+    if onnx or xml:  # OpenVINO requires ONNX
+        f[2] = export_onnx(model, im, file, opset, train, dynamic, simplify)
+    if xml:  # OpenVINO
+        f[3] = export_openvino(model, im, file)
+    if coreml:
+        _, f[4] = export_coreml(model, im, file)
+
+    # TensorFlow Exports
+    if any((saved_model, pb, tflite, edgetpu, tfjs)):
+        if int8 or edgetpu:  # TFLite --int8 bug https://github.com/ultralytics/yolov5/issues/5707
+            check_requirements(('flatbuffers==1.12',))  # required before `import tensorflow`
+        assert not (tflite and tfjs), 'TFLite and TF.js models must be exported separately, please pass only one type.'
+        model, f[5] = export_saved_model(model, im, file, dynamic, tf_nms=nms or agnostic_nms or tfjs,
+                                         agnostic_nms=agnostic_nms or tfjs, topk_per_class=topk_per_class,
+                                         topk_all=topk_all, conf_thres=conf_thres, iou_thres=iou_thres)  # keras model
+        if pb or tfjs:  # pb prerequisite to tfjs
+            f[6] = export_pb(model, im, file)
+        if tflite or edgetpu:
+            f[7] = export_tflite(model, im, file, int8=int8 or edgetpu, data=data, ncalib=100)
+        if edgetpu:
+            f[8] = export_edgetpu(model, im, file)
+        if tfjs:
+            f[9] = export_tfjs(model, im, file)
+
+    # Finish
+    f = [str(x) for x in f if x]  # filter out '' and None
+    if any(f):
+        LOGGER.info(f'\nExport complete ({time.time() - t:.2f}s)'
+                    f"\nResults saved to {colorstr('bold', file.parent.resolve())}"
+                    f"\nDetect:          python detect.py --weights {f[-1]}"
+                    f"\nPyTorch Hub:     model = torch.hub.load('ultralytics/yolov5', 'custom', '{f[-1]}')"
+                    f"\nValidate:        python val.py --weights {f[-1]}"
+                    f"\nVisualize:       https://netron.app")
+    return f  # return list of exported files/dirs
+
+
+def parse_opt():
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--data', type=str, default=ROOT / 'data/coco128.yaml', help='dataset.yaml path')
+    parser.add_argument('--weights', nargs='+', type=str, default=ROOT / 'yolov5s.pt', help='model.pt path(s)')
+    parser.add_argument('--imgsz', '--img', '--img-size', nargs='+', type=int, default=[640, 640], help='image (h, w)')
+    parser.add_argument('--batch-size', type=int, default=1, help='batch size')
+    parser.add_argument('--device', default='cpu', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')
+    parser.add_argument('--half', action='store_true', help='FP16 half-precision export')
+    parser.add_argument('--inplace', action='store_true', help='set YOLOv5 Detect() inplace=True')
+    parser.add_argument('--train', action='store_true', help='model.train() mode')
+    parser.add_argument('--optimize', action='store_true', help='TorchScript: optimize for mobile')
+    parser.add_argument('--int8', action='store_true', help='CoreML/TF INT8 quantization')
+    parser.add_argument('--dynamic', action='store_true', help='ONNX/TF: dynamic axes')
+    parser.add_argument('--simplify', action='store_true', help='ONNX: simplify model')
+    parser.add_argument('--opset', type=int, default=12, help='ONNX: opset version')
+    parser.add_argument('--verbose', action='store_true', help='TensorRT: verbose log')
+    parser.add_argument('--workspace', type=int, default=4, help='TensorRT: workspace size (GB)')
+    parser.add_argument('--nms', action='store_true', help='TF: add NMS to model')
+    parser.add_argument('--agnostic-nms', action='store_true', help='TF: add agnostic NMS to model')
+    parser.add_argument('--topk-per-class', type=int, default=100, help='TF.js NMS: topk per class to keep')
+    parser.add_argument('--topk-all', type=int, default=100, help='TF.js NMS: topk for all classes to keep')
+    parser.add_argument('--iou-thres', type=float, default=0.45, help='TF.js NMS: IoU threshold')
+    parser.add_argument('--conf-thres', type=float, default=0.25, help='TF.js NMS: confidence threshold')
+    parser.add_argument('--include', nargs='+',
+                        default=['torchscript', 'onnx'],
+                        help='torchscript, onnx, openvino, engine, coreml, saved_model, pb, tflite, edgetpu, tfjs')
+    opt = parser.parse_args()
+    print_args(FILE.stem, opt)
+    return opt
+
+
+def main(opt):
+    for opt.weights in (opt.weights if isinstance(opt.weights, list) else [opt.weights]):
+        run(**vars(opt))
+
+
+if __name__ == "__main__":
+    opt = parse_opt()
+    main(opt)

finetune_pruned.py

@@ -0,0 +1,660 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+"""
+Train a YOLOv5 model on a custom dataset.
+
+Models and datasets download automatically from the latest YOLOv5 release.
+Models: https://github.com/ultralytics/yolov5/tree/master/models
+Datasets: https://github.com/ultralytics/yolov5/tree/master/data
+Tutorial: https://github.com/ultralytics/yolov5/wiki/Train-Custom-Data
+
+Usage:
+    $ python path/to/train.py --data coco128.yaml --weights yolov5s.pt --img 640  # from pretrained (RECOMMENDED)
+    $ python path/to/train.py --data coco128.yaml --weights '' --cfg yolov5s.yaml --img 640  # from scratch
+"""
+
+import argparse
+import math
+import os
+import random
+import sys
+import time
+from copy import deepcopy
+from datetime import datetime
+from pathlib import Path
+from models.yolo import ModelPruned
+import numpy as np
+import torch
+import torch.distributed as dist
+import torch.nn as nn
+import yaml
+from torch.cuda import amp
+from torch.nn.parallel import DistributedDataParallel as DDP
+from torch.optim import SGD, Adam, AdamW, lr_scheduler
+from tqdm import tqdm
+from models.common import Bottleneck
+
+FILE = Path(__file__).resolve()
+ROOT = FILE.parents[0]  # YOLOv5 root directory
+if str(ROOT) not in sys.path:
+    sys.path.append(str(ROOT))  # add ROOT to PATH
+ROOT = Path(os.path.relpath(ROOT, Path.cwd()))  # relative
+
+import val  # for end-of-epoch mAP
+from models.experimental import attempt_load
+from models.yolo import Model
+from utils.autoanchor import check_anchors
+from utils.autobatch import check_train_batch_size
+from utils.callbacks import Callbacks
+from utils.datasets import create_dataloader
+from utils.downloads import attempt_download
+from utils.general import (LOGGER, check_dataset, check_file, check_git_status, check_img_size, check_requirements,
+                           check_suffix, check_yaml, colorstr, get_latest_run, increment_path, init_seeds,
+                           intersect_dicts, labels_to_class_weights, labels_to_image_weights, methods, one_cycle,
+                           print_args, print_mutation, strip_optimizer)
+from utils.loggers import Loggers
+from utils.loggers.wandb.wandb_utils import check_wandb_resume
+from utils.loss import ComputeLoss
+from utils.metrics import fitness
+from utils.plots import plot_evolve, plot_labels
+from utils.torch_utils import EarlyStopping, ModelEMA, de_parallel, select_device, torch_distributed_zero_first
+
+LOCAL_RANK = int(os.getenv('LOCAL_RANK', -1))  # https://pytorch.org/docs/stable/elastic/run.html
+RANK = int(os.getenv('RANK', -1))
+WORLD_SIZE = int(os.getenv('WORLD_SIZE', 1))
+
+
+def train(hyp,  # path/to/hyp.yaml or hyp dictionary
+          opt,
+          device,
+          callbacks
+          ):
+    save_dir, epochs, batch_size, weights, single_cls, evolve, data, cfg, resume, noval, nosave, workers, freeze = \
+        Path(opt.save_dir), opt.epochs, opt.batch_size, opt.weights, opt.single_cls, opt.evolve, opt.data, opt.cfg, \
+        opt.resume, opt.noval, opt.nosave, opt.workers, opt.freeze
+
+    # Directories
+    w = save_dir / 'weights'  # weights dir
+    (w.parent if evolve else w).mkdir(parents=True, exist_ok=True)  # make dir
+    last, best = w / 'last.pt', w / 'best.pt'
+
+    # Hyperparameters
+    if isinstance(hyp, str):
+        with open(hyp, errors='ignore') as f:
+            hyp = yaml.safe_load(f)  # load hyps dict
+    LOGGER.info(colorstr('hyperparameters: ') + ', '.join(f'{k}={v}' for k, v in hyp.items()))
+
+    # Save run settings
+    if not evolve:
+        with open(save_dir / 'hyp.yaml', 'w') as f:
+            yaml.safe_dump(hyp, f, sort_keys=False)
+        with open(save_dir / 'opt.yaml', 'w') as f:
+            yaml.safe_dump(vars(opt), f, sort_keys=False)
+
+    # Loggers
+    data_dict = None
+    if RANK in [-1, 0]:
+        loggers = Loggers(save_dir, weights, opt, hyp, LOGGER)  # loggers instance
+        if loggers.wandb:
+            data_dict = loggers.wandb.data_dict
+            if resume:
+                weights, epochs, hyp, batch_size = opt.weights, opt.epochs, opt.hyp, opt.batch_size
+
+        # Register actions
+        for k in methods(loggers):
+            callbacks.register_action(k, callback=getattr(loggers, k))
+
+    # Config
+    plots = not evolve  # create plots
+    cuda = device.type != 'cpu'
+    init_seeds(1 + RANK)
+    with torch_distributed_zero_first(LOCAL_RANK):
+        data_dict = data_dict or check_dataset(data)  # check if None
+    train_path, val_path = data_dict['train'], data_dict['val']
+    nc = 1 if single_cls else int(data_dict['nc'])  # number of classes
+    names = ['item'] if single_cls and len(data_dict['names']) != 1 else data_dict['names']  # class names
+    assert len(names) == nc, f'{len(names)} names found for nc={nc} dataset in {data}'  # check
+    is_coco = isinstance(val_path, str) and val_path.endswith('coco/val2017.txt')  # COCO dataset
+
+    # Model
+    check_suffix(weights, '.pt')  # check weights
+    pretrained = weights.endswith('.pt')
+    if pretrained:
+        with torch_distributed_zero_first(LOCAL_RANK):
+            weights = attempt_download(weights)  # download if not found locally
+        ckpt = torch.load(weights, map_location=device)  # load checkpoint
+        model = ckpt["model"]
+        maskbndict = ckpt['model'].maskbndict
+        model = ModelPruned(maskbndict, ckpt['model'].yaml, ch=3, nc=nc, anchors=hyp.get('anchors')).to(device)  # create
+        exclude = ['anchor'] if (cfg or hyp.get('anchors')) and not resume else []  # exclude keys
+        csd = ckpt['model'].float().state_dict()  # checkpoint state_dict as FP32
+        csd = intersect_dicts(csd, model.state_dict(), exclude=exclude)  # intersect
+        model.load_state_dict(csd, strict=False)  # load
+        LOGGER.info(f'Transferred {len(csd)}/{len(model.state_dict())} items from {weights}')  # report
+    else:
+        LOGGER.info('No pruned weights loaded, please set the right pruned weight path ...')  # report
+        return
+
+    # Freeze
+    freeze = [f'model.{x}.' for x in (freeze if len(freeze) > 1 else range(freeze[0]))]  # layers to freeze
+    for k, v in model.named_parameters():
+        v.requires_grad = True  # train all layers
+        if any(x in k for x in freeze):
+            LOGGER.info(f'freezing {k}')
+            v.requires_grad = False
+
+    # Image size
+    gs = max(int(model.stride.max()), 32)  # grid size (max stride)
+    imgsz = check_img_size(opt.imgsz, gs, floor=gs * 2)  # verify imgsz is gs-multiple
+
+    # Batch size
+    if RANK == -1 and batch_size == -1:  # single-GPU only, estimate best batch size
+        batch_size = check_train_batch_size(model, imgsz)
+        loggers.on_params_update({"batch_size": batch_size})
+
+    # Optimizer
+    nbs = 64  # nominal batch size
+    accumulate = max(round(nbs / batch_size), 1)  # accumulate loss before optimizing
+    hyp['weight_decay'] *= batch_size * accumulate / nbs  # scale weight_decay
+    LOGGER.info(f"Scaled weight_decay = {hyp['weight_decay']}")
+
+    g0, g1, g2 = [], [], []  # optimizer parameter groups
+    for v in model.modules():
+        if hasattr(v, 'bias') and isinstance(v.bias, nn.Parameter):  # bias
+            g2.append(v.bias)
+        if isinstance(v, nn.BatchNorm2d):  # weight (no decay)
+            g0.append(v.weight)
+        elif hasattr(v, 'weight') and isinstance(v.weight, nn.Parameter):  # weight (with decay)
+            g1.append(v.weight)
+
+    if opt.optimizer == 'Adam':
+        optimizer = Adam(g0, lr=hyp['lr0'], betas=(hyp['momentum'], 0.999))  # adjust beta1 to momentum
+    elif opt.optimizer == 'AdamW':
+        optimizer = AdamW(g0, lr=hyp['lr0'], betas=(hyp['momentum'], 0.999))  # adjust beta1 to momentum
+    else:
+        optimizer = SGD(g0, lr=hyp['lr0'], momentum=hyp['momentum'], nesterov=True)
+
+    optimizer.add_param_group({'params': g1, 'weight_decay': hyp['weight_decay']})  # add g1 with weight_decay
+    optimizer.add_param_group({'params': g2})  # add g2 (biases)
+    LOGGER.info(f"{colorstr('optimizer:')} {type(optimizer).__name__} with parameter groups "
+                f"{len(g0)} weight (no decay), {len(g1)} weight, {len(g2)} bias")
+    del g0, g1, g2
+
+    # Scheduler
+    if opt.cos_lr:
+        lf = one_cycle(1, hyp['lrf'], epochs)  # cosine 1->hyp['lrf']
+    else:
+        lf = lambda x: (1 - x / epochs) * (1.0 - hyp['lrf']) + hyp['lrf']  # linear
+    scheduler = lr_scheduler.LambdaLR(optimizer, lr_lambda=lf)  # plot_lr_scheduler(optimizer, scheduler, epochs)
+
+    # EMA
+    ema = ModelEMA(model) if RANK in [-1, 0] else None
+
+    # Resume
+    start_epoch, best_fitness = 0, 0.0
+    
+    # DP mode
+    if cuda and RANK == -1 and torch.cuda.device_count() > 1:
+        LOGGER.warning('WARNING: DP not recommended, use torch.distributed.run for best DDP Multi-GPU results.\n'
+                       'See Multi-GPU Tutorial at https://github.com/ultralytics/yolov5/issues/475 to get started.')
+        model = torch.nn.DataParallel(model)
+
+    # SyncBatchNorm
+    if opt.sync_bn and cuda and RANK != -1:
+        model = torch.nn.SyncBatchNorm.convert_sync_batchnorm(model).to(device)
+        LOGGER.info('Using SyncBatchNorm()')
+
+    # Trainloader
+    train_loader, dataset = create_dataloader(train_path, imgsz, batch_size // WORLD_SIZE, gs, single_cls,
+                                              hyp=hyp, augment=True, cache=None if opt.cache == 'val' else opt.cache,
+                                              rect=opt.rect, rank=LOCAL_RANK, workers=workers,
+                                              image_weights=opt.image_weights, quad=opt.quad,
+                                              prefix=colorstr('train: '), shuffle=True)
+    mlc = int(np.concatenate(dataset.labels, 0)[:, 0].max())  # max label class
+    nb = len(train_loader)  # number of batches
+    assert mlc < nc, f'Label class {mlc} exceeds nc={nc} in {data}. Possible class labels are 0-{nc - 1}'
+
+    # Process 0
+    if RANK in [-1, 0]:
+        val_loader = create_dataloader(val_path, imgsz, batch_size // WORLD_SIZE * 2, gs, single_cls,
+                                       hyp=hyp, cache=None if noval else opt.cache,
+                                       rect=True, rank=-1, workers=workers * 2, pad=0.5,
+                                       prefix=colorstr('val: '))[0]
+
+        if not resume:
+            labels = np.concatenate(dataset.labels, 0)
+            # c = torch.tensor(labels[:, 0])  # classes
+            # cf = torch.bincount(c.long(), minlength=nc) + 1.  # frequency
+            # model._initialize_biases(cf.to(device))
+            if plots:
+                plot_labels(labels, names, save_dir)
+
+            # Anchors
+            if not opt.noautoanchor:
+                check_anchors(dataset, model=model, thr=hyp['anchor_t'], imgsz=imgsz)
+            model.half().float()  # pre-reduce anchor precision
+
+        callbacks.run('on_pretrain_routine_end')
+
+    # DDP mode
+    if cuda and RANK != -1:
+        model = DDP(model, device_ids=[LOCAL_RANK], output_device=LOCAL_RANK)
+
+    # Model attributes
+    nl = de_parallel(model).model[-1].nl  # number of detection layers (to scale hyps)
+    hyp['box'] *= 3 / nl  # scale to layers
+    hyp['cls'] *= nc / 80 * 3 / nl  # scale to classes and layers
+    hyp['obj'] *= (imgsz / 640) ** 2 * 3 / nl  # scale to image size and layers
+    hyp['label_smoothing'] = opt.label_smoothing
+    model.nc = nc  # attach number of classes to model
+    model.hyp = hyp  # attach hyperparameters to model
+    model.class_weights = labels_to_class_weights(dataset.labels, nc).to(device) * nc  # attach class weights
+    model.names = names
+
+    # Start training
+    t0 = time.time()
+    nw = max(round(hyp['warmup_epochs'] * nb), 1000)  # number of warmup iterations, max(3 epochs, 1k iterations)
+    # nw = min(nw, (epochs - start_epoch) / 2 * nb)  # limit warmup to < 1/2 of training
+    last_opt_step = -1
+    maps = np.zeros(nc)  # mAP per class
+    results = (0, 0, 0, 0, 0, 0, 0)  # P, R, mAP@.5, mAP@.5-.95, val_loss(box, obj, cls)
+    scheduler.last_epoch = start_epoch - 1  # do not move
+    scaler = amp.GradScaler(enabled=cuda)
+    stopper = EarlyStopping(patience=opt.patience)
+    compute_loss = ComputeLoss(model)  # init loss class
+    LOGGER.info(f'Image sizes {imgsz} train, {imgsz} val\n'
+                f'Using {train_loader.num_workers * WORLD_SIZE} dataloader workers\n'
+                f"Logging results to {colorstr('bold', save_dir)}\n"
+                f'Starting training for {epochs} epochs...')
+    for epoch in range(start_epoch, epochs):  # epoch ------------------------------------------------------------------
+        model.train()
+
+        ignore_bn_list = []
+        for k, m in model.named_modules():
+            if isinstance(m, Bottleneck):
+                if m.add:
+                    ignore_bn_list.append(k.rsplit(".", 2)[0] + ".cv1.bn")
+                    ignore_bn_list.append(k + '.cv1.bn')
+                    ignore_bn_list.append(k + '.cv2.bn')
+
+
+        # Update image weights (optional, single-GPU only)
+        if opt.image_weights:
+            cw = model.class_weights.cpu().numpy() * (1 - maps) ** 2 / nc  # class weights
+            iw = labels_to_image_weights(dataset.labels, nc=nc, class_weights=cw)  # image weights
+            dataset.indices = random.choices(range(dataset.n), weights=iw, k=dataset.n)  # rand weighted idx
+
+        # Update mosaic border (optional)
+        # b = int(random.uniform(0.25 * imgsz, 0.75 * imgsz + gs) // gs * gs)
+        # dataset.mosaic_border = [b - imgsz, -b]  # height, width borders
+
+        mloss = torch.zeros(3, device=device)  # mean losses
+        if RANK != -1:
+            train_loader.sampler.set_epoch(epoch)
+        pbar = enumerate(train_loader)
+        LOGGER.info(('\n' + '%10s' * 7) % ('Epoch', 'gpu_mem', 'box', 'obj', 'cls', 'labels', 'img_size'))
+        if RANK in [-1, 0]:
+            pbar = tqdm(pbar, total=nb, bar_format='{l_bar}{bar:10}{r_bar}{bar:-10b}')  # progress bar
+        optimizer.zero_grad()
+        for i, (imgs, targets, paths, _) in pbar:  # batch -------------------------------------------------------------
+            ni = i + nb * epoch  # number integrated batches (since train start)
+            imgs = imgs.to(device, non_blocking=True).float() / 255  # uint8 to float32, 0-255 to 0.0-1.0
+
+            # Warmup
+            if ni <= nw:
+                xi = [0, nw]  # x interp
+                # compute_loss.gr = np.interp(ni, xi, [0.0, 1.0])  # iou loss ratio (obj_loss = 1.0 or iou)
+                accumulate = max(1, np.interp(ni, xi, [1, nbs / batch_size]).round())
+                for j, x in enumerate(optimizer.param_groups):
+                    # bias lr falls from 0.1 to lr0, all other lrs rise from 0.0 to lr0
+                    x['lr'] = np.interp(ni, xi, [hyp['warmup_bias_lr'] if j == 2 else 0.0, x['initial_lr'] * lf(epoch)])
+                    if 'momentum' in x:
+                        x['momentum'] = np.interp(ni, xi, [hyp['warmup_momentum'], hyp['momentum']])
+
+            # Multi-scale
+            if opt.multi_scale:
+                sz = random.randrange(imgsz * 0.5, imgsz * 1.5 + gs) // gs * gs  # size
+                sf = sz / max(imgs.shape[2:])  # scale factor
+                if sf != 1:
+                    ns = [math.ceil(x * sf / gs) * gs for x in imgs.shape[2:]]  # new shape (stretched to gs-multiple)
+                    imgs = nn.functional.interpolate(imgs, size=ns, mode='bilinear', align_corners=False)
+
+            # Forward
+            with amp.autocast(enabled=cuda):
+                pred = model(imgs)  # forward
+                loss, loss_items = compute_loss(pred, targets.to(device))  # loss scaled by batch_size
+                if RANK != -1:
+                    loss *= WORLD_SIZE  # gradient averaged between devices in DDP mode
+                if opt.quad:
+                    loss *= 4.
+
+            # Backward
+            scaler.scale(loss).backward()
+
+            # Optimize
+            if ni - last_opt_step >= accumulate:
+                scaler.step(optimizer)  # optimizer.step
+                scaler.update()
+                optimizer.zero_grad()
+                if ema:
+                    ema.update(model)
+                last_opt_step = ni
+
+            # Log
+            if RANK in [-1, 0]:
+                mloss = (mloss * i + loss_items) / (i + 1)  # update mean losses
+                mem = f'{torch.cuda.memory_reserved() / 1E9 if torch.cuda.is_available() else 0:.3g}G'  # (GB)
+                pbar.set_description(('%10s' * 2 + '%10.4g' * 5) % (
+                    f'{epoch}/{epochs - 1}', mem, *mloss, targets.shape[0], imgs.shape[-1]))
+                callbacks.run('on_train_batch_end', ni, model, imgs, targets, paths, plots, opt.sync_bn)
+                if callbacks.stop_training:
+                    return
+            # end batch ------------------------------------------------------------------------------------------------
+
+        # Scheduler
+        lr = [x['lr'] for x in optimizer.param_groups]  # for loggers
+        scheduler.step()
+
+
+        # =============== show bn weights ===================== #
+        module_list = []
+        # module_bias_list = []
+        for i, layer in model.named_modules():
+            if isinstance(layer, nn.BatchNorm2d) and i not in ignore_bn_list:
+                bnw = layer.state_dict()['weight']
+                bnb = layer.state_dict()['bias']
+                module_list.append(bnw)
+                # module_bias_list.append(bnb)
+                # bnw = bnw.sort()
+                # print(f"{i} : {bnw} : ")
+        size_list = [idx.data.shape[0] for idx in module_list]
+
+        bn_weights = torch.zeros(sum(size_list))
+        bnb_weights = torch.zeros(sum(size_list))
+        index = 0
+        for idx, size in enumerate(size_list):
+            bn_weights[index:(index + size)] = module_list[idx].data.abs().clone()
+            # bnb_weights[index:(index + size)] = module_bias_list[idx].data.abs().clone()
+            index += size
+
+
+        if RANK in [-1, 0]:
+            # mAP
+            callbacks.run('on_train_epoch_end', epoch=epoch)
+            ema.update_attr(model, include=['yaml', 'nc', 'hyp', 'names', 'stride', 'class_weights'])
+            final_epoch = (epoch + 1 == epochs) or stopper.possible_stop
+            if not noval or final_epoch:  # Calculate mAP
+                results, maps, _ = val.run(data_dict,
+                                           batch_size=batch_size // WORLD_SIZE * 2,
+                                           imgsz=imgsz,
+                                           model=ema.ema,
+                                           single_cls=single_cls,
+                                           dataloader=val_loader,
+                                           save_dir=save_dir,
+                                           plots=False,
+                                           callbacks=callbacks,
+                                           compute_loss=compute_loss)
+
+            # Update best mAP
+            fi = fitness(np.array(results).reshape(1, -1))  # weighted combination of [P, R, mAP@.5, mAP@.5-.95]
+            if fi > best_fitness:
+                best_fitness = fi
+            log_vals = list(mloss) + list(results) + lr #+ [0]
+            callbacks.run('on_fit_epoch_end', log_vals, epoch, best_fitness, fi)
+            callbacks.run('on_fit_epoch_end_prune', bn_weights.numpy(), epoch)
+
+            # Save model
+            if (not nosave) or (final_epoch and not evolve):  # if save
+                ckpt = {'epoch': epoch,
+                        'best_fitness': best_fitness,
+                        'model': deepcopy(de_parallel(model)).half(),
+                        'ema': deepcopy(ema.ema).half(),
+                        'updates': ema.updates,
+                        'optimizer': optimizer.state_dict(),
+                        'wandb_id': loggers.wandb.wandb_run.id if loggers.wandb else None,
+                        'date': datetime.now().isoformat()}
+
+                # Save last, best and delete
+                torch.save(ckpt, last)
+                if best_fitness == fi:
+                    torch.save(ckpt, best)
+                if (epoch > 0) and (opt.save_period > 0) and (epoch % opt.save_period == 0):
+                    torch.save(ckpt, w / f'epoch{epoch}.pt')
+                del ckpt
+                callbacks.run('on_model_save', last, epoch, final_epoch, best_fitness, fi)
+
+            # Stop Single-GPU
+            if RANK == -1 and stopper(epoch=epoch, fitness=fi):
+                break
+
+            # Stop DDP TODO: known issues shttps://github.com/ultralytics/yolov5/pull/4576
+            # stop = stopper(epoch=epoch, fitness=fi)
+            # if RANK == 0:
+            #    dist.broadcast_object_list([stop], 0)  # broadcast 'stop' to all ranks
+
+        # Stop DPP
+        # with torch_distributed_zero_first(RANK):
+        # if stop:
+        #    break  # must break all DDP ranks
+
+        # end epoch ----------------------------------------------------------------------------------------------------
+    # end training -----------------------------------------------------------------------------------------------------
+    if RANK in [-1, 0]:
+        LOGGER.info(f'\n{epoch - start_epoch + 1} epochs completed in {(time.time() - t0) / 3600:.3f} hours.')
+        for f in last, best:
+            if f.exists():
+                strip_optimizer(f)  # strip optimizers
+                if f is best:
+                    LOGGER.info(f'\nValidating {f}...')
+                    results, _, _ = val.run(data_dict,
+                                            batch_size=batch_size // WORLD_SIZE * 2,
+                                            imgsz=imgsz,
+                                            model=attempt_load(f, device).half(),
+                                            iou_thres=0.65 if is_coco else 0.60,  # best pycocotools results at 0.65
+                                            single_cls=single_cls,
+                                            dataloader=val_loader,
+                                            save_dir=save_dir,
+                                            save_json=is_coco,
+                                            verbose=True,
+                                            plots=True,
+                                            callbacks=callbacks,
+                                            compute_loss=compute_loss)  # val best model with plots
+                    if is_coco:
+                        callbacks.run('on_fit_epoch_end', list(mloss) + list(results) + lr, epoch, best_fitness, fi)
+
+        callbacks.run('on_train_end', last, best, plots, epoch, results)
+        LOGGER.info(f"Results saved to {colorstr('bold', save_dir)}")
+
+    torch.cuda.empty_cache()
+    return results
+
+
+def parse_opt(known=False):
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--weights', type=str, default=ROOT / 'VOC2007_wm/prune/exp4/weights/pruned_model.pt', help='initial weights path')
+    parser.add_argument('--cfg', type=str, default='', help='model.yaml path')
+    parser.add_argument('--data', type=str, default=ROOT / 'data/VOC.yaml', help='dataset.yaml path')
+    parser.add_argument('--hyp', type=str, default=ROOT / 'data/hyps/hyp.scratch-low.yaml', help='hyperparameters path')
+    parser.add_argument('--epochs', type=int, default=100)
+    parser.add_argument('--batch-size', type=int, default=16, help='total batch size for all GPUs, -1 for autobatch')
+    parser.add_argument('--imgsz', '--img', '--img-size', type=int, default=640, help='train, val image size (pixels)')
+    parser.add_argument('--rect', action='store_true', help='rectangular training')
+    parser.add_argument('--resume', nargs='?', const=True, default=False, help='resume most recent training')
+    parser.add_argument('--nosave', action='store_true', help='only save final checkpoint')
+    parser.add_argument('--noval', action='store_true', help='only validate final epoch')
+    parser.add_argument('--noautoanchor', action='store_true', help='disable AutoAnchor')
+    parser.add_argument('--evolve', type=int, nargs='?', const=300, help='evolve hyperparameters for x generations')
+    parser.add_argument('--bucket', type=str, default='', help='gsutil bucket')
+    parser.add_argument('--cache', type=str, nargs='?', const='ram', help='--cache images in "ram" (default) or "disk"')
+    parser.add_argument('--image-weights', action='store_true', help='use weighted image selection for training')
+    parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')
+    parser.add_argument('--multi-scale', action='store_true', help='vary img-size +/- 50%%')
+    parser.add_argument('--single-cls', action='store_true', help='train multi-class data as single-class')
+    parser.add_argument('--optimizer', type=str, choices=['SGD', 'Adam', 'AdamW'], default='SGD', help='optimizer')
+    parser.add_argument('--sync-bn', action='store_true', help='use SyncBatchNorm, only available in DDP mode')
+    parser.add_argument('--workers', type=int, default=8, help='max dataloader workers (per RANK in DDP mode)')
+    parser.add_argument('--project', default=ROOT / 'VOC2007_wm/finetune_pruned', help='save to project/name')
+    parser.add_argument('--name', default='exp', help='save to project/name')
+    parser.add_argument('--exist-ok', action='store_true', help='existing project/name ok, do not increment')
+    parser.add_argument('--quad', action='store_true', help='quad dataloader')
+    parser.add_argument('--cos-lr', action='store_true', help='cosine LR scheduler')
+    parser.add_argument('--label-smoothing', type=float, default=0.0, help='Label smoothing epsilon')
+    parser.add_argument('--patience', type=int, default=100, help='EarlyStopping patience (epochs without improvement)')
+    parser.add_argument('--freeze', nargs='+', type=int, default=[0], help='Freeze layers: backbone=10, first3=0 1 2')
+    parser.add_argument('--save-period', type=int, default=-1, help='Save checkpoint every x epochs (disabled if < 1)')
+    parser.add_argument('--local_rank', type=int, default=-1, help='DDP parameter, do not modify')
+
+    # Weights & Biases arguments
+    parser.add_argument('--entity', default=None, help='W&B: Entity')
+    parser.add_argument('--upload_dataset', nargs='?', const=True, default=False, help='W&B: Upload data, "val" option')
+    parser.add_argument('--bbox_interval', type=int, default=-1, help='W&B: Set bounding-box image logging interval')
+    parser.add_argument('--artifact_alias', type=str, default='latest', help='W&B: Version of dataset artifact to use')
+
+    opt = parser.parse_known_args()[0] if known else parser.parse_args()
+    return opt
+
+
+def main(opt, callbacks=Callbacks()):
+    # Checks
+    if RANK in [-1, 0]:
+        print_args(FILE.stem, opt)
+        check_git_status()
+        check_requirements(exclude=['thop'])
+
+    # Resume
+    if opt.resume and not check_wandb_resume(opt) and not opt.evolve:  # resume an interrupted run
+        ckpt = opt.resume if isinstance(opt.resume, str) else get_latest_run()  # specified or most recent path
+        assert os.path.isfile(ckpt), 'ERROR: --resume checkpoint does not exist'
+        with open(Path(ckpt).parent.parent / 'opt.yaml', errors='ignore') as f:
+            opt = argparse.Namespace(**yaml.safe_load(f))  # replace
+        opt.cfg, opt.weights, opt.resume = '', ckpt, True  # reinstate
+        LOGGER.info(f'Resuming training from {ckpt}')
+    else:
+        opt.data, opt.cfg, opt.hyp, opt.weights, opt.project = \
+            check_file(opt.data), check_yaml(opt.cfg), check_yaml(opt.hyp), str(opt.weights), str(opt.project)  # checks
+        assert len(opt.cfg) or len(opt.weights), 'either --cfg or --weights must be specified'
+        if opt.evolve:
+            if opt.project == str(ROOT / 'runs/train'):  # if default project name, rename to runs/evolve
+                opt.project = str(ROOT / 'runs/evolve')
+            opt.exist_ok, opt.resume = opt.resume, False  # pass resume to exist_ok and disable resume
+        opt.save_dir = str(increment_path(Path(opt.project) / opt.name, exist_ok=opt.exist_ok))
+
+    # DDP mode
+    device = select_device(opt.device, batch_size=opt.batch_size)
+    if LOCAL_RANK != -1:
+        msg = 'is not compatible with YOLOv5 Multi-GPU DDP training'
+        assert not opt.image_weights, f'--image-weights {msg}'
+        assert not opt.evolve, f'--evolve {msg}'
+        assert opt.batch_size != -1, f'AutoBatch with --batch-size -1 {msg}, please pass a valid --batch-size'
+        assert opt.batch_size % WORLD_SIZE == 0, f'--batch-size {opt.batch_size} must be multiple of WORLD_SIZE'
+        assert torch.cuda.device_count() > LOCAL_RANK, 'insufficient CUDA devices for DDP command'
+        torch.cuda.set_device(LOCAL_RANK)
+        device = torch.device('cuda', LOCAL_RANK)
+        dist.init_process_group(backend="nccl" if dist.is_nccl_available() else "gloo")
+
+    # Train
+    if not opt.evolve:
+        train(opt.hyp, opt, device, callbacks)
+        if WORLD_SIZE > 1 and RANK == 0:
+            LOGGER.info('Destroying process group... ')
+            dist.destroy_process_group()
+
+    # Evolve hyperparameters (optional)
+    else:
+        # Hyperparameter evolution metadata (mutation scale 0-1, lower_limit, upper_limit)
+        meta = {'lr0': (1, 1e-5, 1e-1),  # initial learning rate (SGD=1E-2, Adam=1E-3)
+                'lrf': (1, 0.01, 1.0),  # final OneCycleLR learning rate (lr0 * lrf)
+                'momentum': (0.3, 0.6, 0.98),  # SGD momentum/Adam beta1
+                'weight_decay': (1, 0.0, 0.001),  # optimizer weight decay
+                'warmup_epochs': (1, 0.0, 5.0),  # warmup epochs (fractions ok)
+                'warmup_momentum': (1, 0.0, 0.95),  # warmup initial momentum
+                'warmup_bias_lr': (1, 0.0, 0.2),  # warmup initial bias lr
+                'box': (1, 0.02, 0.2),  # box loss gain
+                'cls': (1, 0.2, 4.0),  # cls loss gain
+                'cls_pw': (1, 0.5, 2.0),  # cls BCELoss positive_weight
+                'obj': (1, 0.2, 4.0),  # obj loss gain (scale with pixels)
+                'obj_pw': (1, 0.5, 2.0),  # obj BCELoss positive_weight
+                'iou_t': (0, 0.1, 0.7),  # IoU training threshold
+                'anchor_t': (1, 2.0, 8.0),  # anchor-multiple threshold
+                'anchors': (2, 2.0, 10.0),  # anchors per output grid (0 to ignore)
+                'fl_gamma': (0, 0.0, 2.0),  # focal loss gamma (efficientDet default gamma=1.5)
+                'hsv_h': (1, 0.0, 0.1),  # image HSV-Hue augmentation (fraction)
+                'hsv_s': (1, 0.0, 0.9),  # image HSV-Saturation augmentation (fraction)
+                'hsv_v': (1, 0.0, 0.9),  # image HSV-Value augmentation (fraction)
+                'degrees': (1, 0.0, 45.0),  # image rotation (+/- deg)
+                'translate': (1, 0.0, 0.9),  # image translation (+/- fraction)
+                'scale': (1, 0.0, 0.9),  # image scale (+/- gain)
+                'shear': (1, 0.0, 10.0),  # image shear (+/- deg)
+                'perspective': (0, 0.0, 0.001),  # image perspective (+/- fraction), range 0-0.001
+                'flipud': (1, 0.0, 1.0),  # image flip up-down (probability)
+                'fliplr': (0, 0.0, 1.0),  # image flip left-right (probability)
+                'mosaic': (1, 0.0, 1.0),  # image mixup (probability)
+                'mixup': (1, 0.0, 1.0),  # image mixup (probability)
+                'copy_paste': (1, 0.0, 1.0)}  # segment copy-paste (probability)
+
+        with open(opt.hyp, errors='ignore') as f:
+            hyp = yaml.safe_load(f)  # load hyps dict
+            if 'anchors' not in hyp:  # anchors commented in hyp.yaml
+                hyp['anchors'] = 3
+        opt.noval, opt.nosave, save_dir = True, True, Path(opt.save_dir)  # only val/save final epoch
+        # ei = [isinstance(x, (int, float)) for x in hyp.values()]  # evolvable indices
+        evolve_yaml, evolve_csv = save_dir / 'hyp_evolve.yaml', save_dir / 'evolve.csv'
+        if opt.bucket:
+            os.system(f'gsutil cp gs://{opt.bucket}/evolve.csv {evolve_csv}')  # download evolve.csv if exists
+
+        for _ in range(opt.evolve):  # generations to evolve
+            if evolve_csv.exists():  # if evolve.csv exists: select best hyps and mutate
+                # Select parent(s)
+                parent = 'single'  # parent selection method: 'single' or 'weighted'
+                x = np.loadtxt(evolve_csv, ndmin=2, delimiter=',', skiprows=1)
+                n = min(5, len(x))  # number of previous results to consider
+                x = x[np.argsort(-fitness(x))][:n]  # top n mutations
+                w = fitness(x) - fitness(x).min() + 1E-6  # weights (sum > 0)
+                if parent == 'single' or len(x) == 1:
+                    # x = x[random.randint(0, n - 1)]  # random selection
+                    x = x[random.choices(range(n), weights=w)[0]]  # weighted selection
+                elif parent == 'weighted':
+                    x = (x * w.reshape(n, 1)).sum(0) / w.sum()  # weighted combination
+
+                # Mutate
+                mp, s = 0.8, 0.2  # mutation probability, sigma
+                npr = np.random
+                npr.seed(int(time.time()))
+                g = np.array([meta[k][0] for k in hyp.keys()])  # gains 0-1
+                ng = len(meta)
+                v = np.ones(ng)
+                while all(v == 1):  # mutate until a change occurs (prevent duplicates)
+                    v = (g * (npr.random(ng) < mp) * npr.randn(ng) * npr.random() * s + 1).clip(0.3, 3.0)
+                for i, k in enumerate(hyp.keys()):  # plt.hist(v.ravel(), 300)
+                    hyp[k] = float(x[i + 7] * v[i])  # mutate
+
+            # Constrain to limits
+            for k, v in meta.items():
+                hyp[k] = max(hyp[k], v[1])  # lower limit
+                hyp[k] = min(hyp[k], v[2])  # upper limit
+                hyp[k] = round(hyp[k], 5)  # significant digits
+
+            # Train mutation
+            results = train(hyp.copy(), opt, device, callbacks)
+            callbacks = Callbacks()
+            # Write mutation results
+            print_mutation(results, hyp.copy(), save_dir, opt.bucket)
+
+        # Plot results
+        plot_evolve(evolve_csv)
+        LOGGER.info(f'Hyperparameter evolution finished {opt.evolve} generations\n'
+                    f"Results saved to {colorstr('bold', save_dir)}\n"
+                    f'Usage example: $ python train.py --hyp {evolve_yaml}')
+
+
+def run(**kwargs):
+    # Usage: import train; train.run(data='coco128.yaml', imgsz=320, weights='yolov5m.pt')
+    opt = parse_opt(True)
+    for k, v in kwargs.items():
+        setattr(opt, k, v)
+    main(opt)
+    return opt
+
+
+if __name__ == "__main__":
+    opt = parse_opt()
+    main(opt)

hubconf.py

@@ -0,0 +1,143 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+"""
+PyTorch Hub models https://pytorch.org/hub/ultralytics_yolov5/
+
+Usage:
+    import torch
+    model = torch.hub.load('ultralytics/yolov5', 'yolov5s')
+    model = torch.hub.load('ultralytics/yolov5:master', 'custom', 'path/to/yolov5s.onnx')  # file from branch
+"""
+
+import torch
+
+
+def _create(name, pretrained=True, channels=3, classes=80, autoshape=True, verbose=True, device=None):
+    """Creates or loads a YOLOv5 model
+
+    Arguments:
+        name (str): model name 'yolov5s' or path 'path/to/best.pt'
+        pretrained (bool): load pretrained weights into the model
+        channels (int): number of input channels
+        classes (int): number of model classes
+        autoshape (bool): apply YOLOv5 .autoshape() wrapper to model
+        verbose (bool): print all information to screen
+        device (str, torch.device, None): device to use for model parameters
+
+    Returns:
+        YOLOv5 model
+    """
+    from pathlib import Path
+
+    from models.common import AutoShape, DetectMultiBackend
+    from models.yolo import Model
+    from utils.downloads import attempt_download
+    from utils.general import LOGGER, check_requirements, intersect_dicts, logging
+    from utils.torch_utils import select_device
+
+    if not verbose:
+        LOGGER.setLevel(logging.WARNING)
+    check_requirements(exclude=('tensorboard', 'thop', 'opencv-python'))
+    name = Path(name)
+    path = name.with_suffix('.pt') if name.suffix == '' else name  # checkpoint path
+    try:
+        device = select_device(('0' if torch.cuda.is_available() else 'cpu') if device is None else device)
+
+        if pretrained and channels == 3 and classes == 80:
+            model = DetectMultiBackend(path, device=device)  # download/load FP32 model
+            # model = models.experimental.attempt_load(path, map_location=device)  # download/load FP32 model
+        else:
+            cfg = list((Path(__file__).parent / 'models').rglob(f'{path.stem}.yaml'))[0]  # model.yaml path
+            model = Model(cfg, channels, classes)  # create model
+            if pretrained:
+                ckpt = torch.load(attempt_download(path), map_location=device)  # load
+                csd = ckpt['model'].float().state_dict()  # checkpoint state_dict as FP32
+                csd = intersect_dicts(csd, model.state_dict(), exclude=['anchors'])  # intersect
+                model.load_state_dict(csd, strict=False)  # load
+                if len(ckpt['model'].names) == classes:
+                    model.names = ckpt['model'].names  # set class names attribute
+        if autoshape:
+            model = AutoShape(model)  # for file/URI/PIL/cv2/np inputs and NMS
+        return model.to(device)
+
+    except Exception as e:
+        help_url = 'https://github.com/ultralytics/yolov5/issues/36'
+        s = f'{e}. Cache may be out of date, try `force_reload=True` or see {help_url} for help.'
+        raise Exception(s) from e
+
+
+def custom(path='path/to/model.pt', autoshape=True, verbose=True, device=None):
+    # YOLOv5 custom or local model
+    return _create(path, autoshape=autoshape, verbose=verbose, device=device)
+
+
+def yolov5n(pretrained=True, channels=3, classes=80, autoshape=True, verbose=True, device=None):
+    # YOLOv5-nano model https://github.com/ultralytics/yolov5
+    return _create('yolov5n', pretrained, channels, classes, autoshape, verbose, device)
+
+
+def yolov5s(pretrained=True, channels=3, classes=80, autoshape=True, verbose=True, device=None):
+    # YOLOv5-small model https://github.com/ultralytics/yolov5
+    return _create('yolov5s', pretrained, channels, classes, autoshape, verbose, device)
+
+
+def yolov5m(pretrained=True, channels=3, classes=80, autoshape=True, verbose=True, device=None):
+    # YOLOv5-medium model https://github.com/ultralytics/yolov5
+    return _create('yolov5m', pretrained, channels, classes, autoshape, verbose, device)
+
+
+def yolov5l(pretrained=True, channels=3, classes=80, autoshape=True, verbose=True, device=None):
+    # YOLOv5-large model https://github.com/ultralytics/yolov5
+    return _create('yolov5l', pretrained, channels, classes, autoshape, verbose, device)
+
+
+def yolov5x(pretrained=True, channels=3, classes=80, autoshape=True, verbose=True, device=None):
+    # YOLOv5-xlarge model https://github.com/ultralytics/yolov5
+    return _create('yolov5x', pretrained, channels, classes, autoshape, verbose, device)
+
+
+def yolov5n6(pretrained=True, channels=3, classes=80, autoshape=True, verbose=True, device=None):
+    # YOLOv5-nano-P6 model https://github.com/ultralytics/yolov5
+    return _create('yolov5n6', pretrained, channels, classes, autoshape, verbose, device)
+
+
+def yolov5s6(pretrained=True, channels=3, classes=80, autoshape=True, verbose=True, device=None):
+    # YOLOv5-small-P6 model https://github.com/ultralytics/yolov5
+    return _create('yolov5s6', pretrained, channels, classes, autoshape, verbose, device)
+
+
+def yolov5m6(pretrained=True, channels=3, classes=80, autoshape=True, verbose=True, device=None):
+    # YOLOv5-medium-P6 model https://github.com/ultralytics/yolov5
+    return _create('yolov5m6', pretrained, channels, classes, autoshape, verbose, device)
+
+
+def yolov5l6(pretrained=True, channels=3, classes=80, autoshape=True, verbose=True, device=None):
+    # YOLOv5-large-P6 model https://github.com/ultralytics/yolov5
+    return _create('yolov5l6', pretrained, channels, classes, autoshape, verbose, device)
+
+
+def yolov5x6(pretrained=True, channels=3, classes=80, autoshape=True, verbose=True, device=None):
+    # YOLOv5-xlarge-P6 model https://github.com/ultralytics/yolov5
+    return _create('yolov5x6', pretrained, channels, classes, autoshape, verbose, device)
+
+
+if __name__ == '__main__':
+    model = _create(name='yolov5s', pretrained=True, channels=3, classes=80, autoshape=True, verbose=True)  # pretrained
+    # model = custom(path='path/to/model.pt')  # custom
+
+    # Verify inference
+    from pathlib import Path
+
+    import cv2
+    import numpy as np
+    from PIL import Image
+
+    imgs = ['data/images/zidane.jpg',  # filename
+            Path('data/images/zidane.jpg'),  # Path
+            'https://ultralytics.com/images/zidane.jpg',  # URI
+            cv2.imread('data/images/bus.jpg')[:, :, ::-1],  # OpenCV
+            Image.open('data/images/bus.jpg'),  # PIL
+            np.zeros((320, 640, 3))]  # numpy
+
+    results = model(imgs, size=320)  # batched inference
+    results.print()
+    results.save()

models/common.py

@@ -0,0 +1,871 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+"""
+Common modules
+"""
+
+import json
+import math
+import platform
+import warnings
+from collections import OrderedDict, namedtuple
+from copy import copy
+from pathlib import Path
+
+import cv2
+import numpy as np
+import pandas as pd
+import requests
+import torch
+import torch.nn as nn
+import yaml
+from PIL import Image
+from torch.cuda import amp
+
+from utils.datasets import exif_transpose, letterbox
+from utils.general import (LOGGER, check_requirements, check_suffix, check_version, colorstr, increment_path,
+                           make_divisible, non_max_suppression, scale_coords, xywh2xyxy, xyxy2xywh)
+from utils.plots import Annotator, colors, save_one_box
+from utils.torch_utils import copy_attr, time_sync
+
+
+def autopad(k, p=None):  # kernel, padding
+    # Pad to 'same'
+    if p is None:
+        p = k // 2 if isinstance(k, int) else [x // 2 for x in k]  # auto-pad
+    return p
+
+
+class Conv(nn.Module):
+    # Standard convolution
+    def __init__(self, c1, c2, k=1, s=1, p=None, g=1, act=True):  # ch_in, ch_out, kernel, stride, padding, groups
+        super().__init__()
+        self.conv = nn.Conv2d(c1, c2, k, s, autopad(k, p), groups=g, bias=False)
+        self.bn = nn.BatchNorm2d(c2)
+        self.act = nn.SiLU() if act is True else (act if isinstance(act, nn.Module) else nn.Identity())
+
+    def forward(self, x):
+        return self.act(self.bn(self.conv(x)))
+
+    def forward_fuse(self, x):
+        return self.act(self.conv(x))
+
+
+class DWConv(Conv):
+    # Depth-wise convolution class
+    def __init__(self, c1, c2, k=1, s=1, act=True):  # ch_in, ch_out, kernel, stride, padding, groups
+        super().__init__(c1, c2, k, s, g=math.gcd(c1, c2), act=act)
+
+
+class TransformerLayer(nn.Module):
+    # Transformer layer https://arxiv.org/abs/2010.11929 (LayerNorm layers removed for better performance)
+    def __init__(self, c, num_heads):
+        super().__init__()
+        self.q = nn.Linear(c, c, bias=False)
+        self.k = nn.Linear(c, c, bias=False)
+        self.v = nn.Linear(c, c, bias=False)
+        self.ma = nn.MultiheadAttention(embed_dim=c, num_heads=num_heads)
+        self.fc1 = nn.Linear(c, c, bias=False)
+        self.fc2 = nn.Linear(c, c, bias=False)
+
+    def forward(self, x):
+        x = self.ma(self.q(x), self.k(x), self.v(x))[0] + x
+        x = self.fc2(self.fc1(x)) + x
+        return x
+
+
+class TransformerBlock(nn.Module):
+    # Vision Transformer https://arxiv.org/abs/2010.11929
+    def __init__(self, c1, c2, num_heads, num_layers):
+        super().__init__()
+        self.conv = None
+        if c1 != c2:
+            self.conv = Conv(c1, c2)
+        self.linear = nn.Linear(c2, c2)  # learnable position embedding
+        self.tr = nn.Sequential(*(TransformerLayer(c2, num_heads) for _ in range(num_layers)))
+        self.c2 = c2
+
+    def forward(self, x):
+        if self.conv is not None:
+            x = self.conv(x)
+        b, _, w, h = x.shape
+        p = x.flatten(2).permute(2, 0, 1)
+        return self.tr(p + self.linear(p)).permute(1, 2, 0).reshape(b, self.c2, w, h)
+
+
+class Bottleneck(nn.Module):
+    # Standard bottleneck
+    def __init__(self, c1, c2, shortcut=True, g=1, e=0.5):  # ch_in, ch_out, shortcut, groups, expansion
+        super().__init__()
+        c_ = int(c2 * e)  # hidden channels
+        self.cv1 = Conv(c1, c_, 1, 1)
+        self.cv2 = Conv(c_, c2, 3, 1, g=g)
+        self.add = shortcut and c1 == c2
+
+    def forward(self, x):
+        return x + self.cv2(self.cv1(x)) if self.add else self.cv2(self.cv1(x))
+
+
+class BottleneckCSP(nn.Module):
+    # CSP Bottleneck https://github.com/WongKinYiu/CrossStagePartialNetworks
+    def __init__(self, c1, c2, n=1, shortcut=True, g=1, e=0.5):  # ch_in, ch_out, number, shortcut, groups, expansion
+        super().__init__()
+        c_ = int(c2 * e)  # hidden channels
+        self.cv1 = Conv(c1, c_, 1, 1)
+        self.cv2 = nn.Conv2d(c1, c_, 1, 1, bias=False)
+        self.cv3 = nn.Conv2d(c_, c_, 1, 1, bias=False)
+        self.cv4 = Conv(2 * c_, c2, 1, 1)
+        self.bn = nn.BatchNorm2d(2 * c_)  # applied to cat(cv2, cv3)
+        self.act = nn.SiLU()
+        self.m = nn.Sequential(*(Bottleneck(c_, c_, shortcut, g, e=1.0) for _ in range(n)))
+
+    def forward(self, x):
+        y1 = self.cv3(self.m(self.cv1(x)))
+        y2 = self.cv2(x)
+        return self.cv4(self.act(self.bn(torch.cat((y1, y2), dim=1))))
+
+
+class C3(nn.Module):
+    # CSP Bottleneck with 3 convolutions
+    def __init__(self, c1, c2, n=1, shortcut=True, g=1, e=0.5):  # ch_in, ch_out, number, shortcut, groups, expansion
+        super().__init__()
+        c_ = int(c2 * e)  # hidden channels
+        self.cv1 = Conv(c1, c_, 1, 1)
+        self.cv2 = Conv(c1, c_, 1, 1)
+        self.cv3 = Conv(2 * c_, c2, 1)  # act=FReLU(c2)
+        self.m = nn.Sequential(*(Bottleneck(c_, c_, shortcut, g, e=1.0) for _ in range(n)))
+        # self.m = nn.Sequential(*[CrossConv(c_, c_, 3, 1, g, 1.0, shortcut) for _ in range(n)])
+
+    def forward(self, x):
+        return self.cv3(torch.cat((self.m(self.cv1(x)), self.cv2(x)), dim=1))
+
+
+class C3TR(C3):
+    # C3 module with TransformerBlock()
+    def __init__(self, c1, c2, n=1, shortcut=True, g=1, e=0.5):
+        super().__init__(c1, c2, n, shortcut, g, e)
+        c_ = int(c2 * e)
+        self.m = TransformerBlock(c_, c_, 4, n)
+
+
+class C3SPP(C3):
+    # C3 module with SPP()
+    def __init__(self, c1, c2, k=(5, 9, 13), n=1, shortcut=True, g=1, e=0.5):
+        super().__init__(c1, c2, n, shortcut, g, e)
+        c_ = int(c2 * e)
+        self.m = SPP(c_, c_, k)
+
+
+class C3Ghost(C3):
+    # C3 module with GhostBottleneck()
+    def __init__(self, c1, c2, n=1, shortcut=True, g=1, e=0.5):
+        super().__init__(c1, c2, n, shortcut, g, e)
+        c_ = int(c2 * e)  # hidden channels
+        self.m = nn.Sequential(*(GhostBottleneck(c_, c_) for _ in range(n)))
+
+
+class SPP(nn.Module):
+    # Spatial Pyramid Pooling (SPP) layer https://arxiv.org/abs/1406.4729
+    def __init__(self, c1, c2, k=(5, 9, 13)):
+        super().__init__()
+        c_ = c1 // 2  # hidden channels
+        self.cv1 = Conv(c1, c_, 1, 1)
+        self.cv2 = Conv(c_ * (len(k) + 1), c2, 1, 1)
+        self.m = nn.ModuleList([nn.MaxPool2d(kernel_size=x, stride=1, padding=x // 2) for x in k])
+
+    def forward(self, x):
+        x = self.cv1(x)
+        with warnings.catch_warnings():
+            warnings.simplefilter('ignore')  # suppress torch 1.9.0 max_pool2d() warning
+            return self.cv2(torch.cat([x] + [m(x) for m in self.m], 1))
+
+
+class SPPF(nn.Module):
+    # Spatial Pyramid Pooling - Fast (SPPF) layer for YOLOv5 by Glenn Jocher
+    def __init__(self, c1, c2, k=5):  # equivalent to SPP(k=(5, 9, 13))
+        super().__init__()
+        c_ = c1 // 2  # hidden channels
+        self.cv1 = Conv(c1, c_, 1, 1)
+        self.cv2 = Conv(c_ * 4, c2, 1, 1)
+        self.m = nn.MaxPool2d(kernel_size=k, stride=1, padding=k // 2)
+
+    def forward(self, x):
+        x = self.cv1(x)
+        with warnings.catch_warnings():
+            warnings.simplefilter('ignore')  # suppress torch 1.9.0 max_pool2d() warning
+            y1 = self.m(x)
+            y2 = self.m(y1)
+            return self.cv2(torch.cat([x, y1, y2, self.m(y2)], 1))
+
+
+class Focus(nn.Module):
+    # Focus wh information into c-space
+    def __init__(self, c1, c2, k=1, s=1, p=None, g=1, act=True):  # ch_in, ch_out, kernel, stride, padding, groups
+        super().__init__()
+        self.conv = Conv(c1 * 4, c2, k, s, p, g, act)
+        # self.contract = Contract(gain=2)
+
+    def forward(self, x):  # x(b,c,w,h) -> y(b,4c,w/2,h/2)
+        return self.conv(torch.cat([x[..., ::2, ::2], x[..., 1::2, ::2], x[..., ::2, 1::2], x[..., 1::2, 1::2]], 1))
+        # return self.conv(self.contract(x))
+
+
+class GhostConv(nn.Module):
+    # Ghost Convolution https://github.com/huawei-noah/ghostnet
+    def __init__(self, c1, c2, k=1, s=1, g=1, act=True):  # ch_in, ch_out, kernel, stride, groups
+        super().__init__()
+        c_ = c2 // 2  # hidden channels
+        self.cv1 = Conv(c1, c_, k, s, None, g, act)
+        self.cv2 = Conv(c_, c_, 5, 1, None, c_, act)
+
+    def forward(self, x):
+        y = self.cv1(x)
+        return torch.cat([y, self.cv2(y)], 1)
+
+
+class GhostBottleneck(nn.Module):
+    # Ghost Bottleneck https://github.com/huawei-noah/ghostnet
+    def __init__(self, c1, c2, k=3, s=1):  # ch_in, ch_out, kernel, stride
+        super().__init__()
+        c_ = c2 // 2
+        self.conv = nn.Sequential(GhostConv(c1, c_, 1, 1),  # pw
+                                  DWConv(c_, c_, k, s, act=False) if s == 2 else nn.Identity(),  # dw
+                                  GhostConv(c_, c2, 1, 1, act=False))  # pw-linear
+        self.shortcut = nn.Sequential(DWConv(c1, c1, k, s, act=False),
+                                      Conv(c1, c2, 1, 1, act=False)) if s == 2 else nn.Identity()
+
+    def forward(self, x):
+        return self.conv(x) + self.shortcut(x)
+
+
+class Contract(nn.Module):
+    # Contract width-height into channels, i.e. x(1,64,80,80) to x(1,256,40,40)
+    def __init__(self, gain=2):
+        super().__init__()
+        self.gain = gain
+
+    def forward(self, x):
+        b, c, h, w = x.size()  # assert (h / s == 0) and (W / s == 0), 'Indivisible gain'
+        s = self.gain
+        x = x.view(b, c, h // s, s, w // s, s)  # x(1,64,40,2,40,2)
+        x = x.permute(0, 3, 5, 1, 2, 4).contiguous()  # x(1,2,2,64,40,40)
+        return x.view(b, c * s * s, h // s, w // s)  # x(1,256,40,40)
+
+
+class Expand(nn.Module):
+    # Expand channels into width-height, i.e. x(1,64,80,80) to x(1,16,160,160)
+    def __init__(self, gain=2):
+        super().__init__()
+        self.gain = gain
+
+    def forward(self, x):
+        b, c, h, w = x.size()  # assert C / s ** 2 == 0, 'Indivisible gain'
+        s = self.gain
+        x = x.view(b, s, s, c // s ** 2, h, w)  # x(1,2,2,16,80,80)
+        x = x.permute(0, 3, 4, 1, 5, 2).contiguous()  # x(1,16,80,2,80,2)
+        return x.view(b, c // s ** 2, h * s, w * s)  # x(1,16,160,160)
+
+
+class Concat(nn.Module):
+    # Concatenate a list of tensors along dimension
+    def __init__(self, dimension=1):
+        super().__init__()
+        self.d = dimension
+
+    def forward(self, x):
+        return torch.cat(x, self.d)
+
+
+class DetectMultiBackend(nn.Module):
+    # YOLOv5 MultiBackend class for python inference on various backends
+    def __init__(self, weights='yolov5s.pt', device=None, dnn=False, data=None):
+        # Usage:
+        #   PyTorch:              weights = *.pt
+        #   TorchScript:                    *.torchscript
+        #   ONNX Runtime:                   *.onnx
+        #   ONNX OpenCV DNN:                *.onnx with --dnn
+        #   OpenVINO:                       *.xml
+        #   CoreML:                         *.mlmodel
+        #   TensorRT:                       *.engine
+        #   TensorFlow SavedModel:          *_saved_model
+        #   TensorFlow GraphDef:            *.pb
+        #   TensorFlow Lite:                *.tflite
+        #   TensorFlow Edge TPU:            *_edgetpu.tflite
+        from models.experimental import attempt_download, attempt_load  # scoped to avoid circular import
+
+        super().__init__()
+        w = str(weights[0] if isinstance(weights, list) else weights)
+        pt, jit, onnx, xml, engine, coreml, saved_model, pb, tflite, edgetpu, tfjs = self.model_type(w)  # get backend
+        stride, names = 64, [f'class{i}' for i in range(1000)]  # assign defaults
+        w = attempt_download(w)  # download if not local
+        if data:  # data.yaml path (optional)
+            with open(data, errors='ignore') as f:
+                names = yaml.safe_load(f)['names']  # class names
+
+        if pt:  # PyTorch
+            model = attempt_load(weights if isinstance(weights, list) else w, map_location=device)
+            stride = max(int(model.stride.max()), 32)  # model stride
+            names = model.module.names if hasattr(model, 'module') else model.names  # get class names
+            self.model = model  # explicitly assign for to(), cpu(), cuda(), half()
+        elif jit:  # TorchScript
+            LOGGER.info(f'Loading {w} for TorchScript inference...')
+            extra_files = {'config.txt': ''}  # model metadata
+            model = torch.jit.load(w, _extra_files=extra_files)
+            if extra_files['config.txt']:
+                d = json.loads(extra_files['config.txt'])  # extra_files dict
+                stride, names = int(d['stride']), d['names']
+        elif dnn:  # ONNX OpenCV DNN
+            LOGGER.info(f'Loading {w} for ONNX OpenCV DNN inference...')
+            check_requirements(('opencv-python>=4.5.4',))
+            net = cv2.dnn.readNetFromONNX(w)
+        elif onnx:  # ONNX Runtime
+            LOGGER.info(f'Loading {w} for ONNX Runtime inference...')
+            cuda = torch.cuda.is_available()
+            check_requirements(('onnx', 'onnxruntime-gpu' if cuda else 'onnxruntime'))
+            import onnxruntime
+            providers = ['CUDAExecutionProvider', 'CPUExecutionProvider'] if cuda else ['CPUExecutionProvider']
+            session = onnxruntime.InferenceSession(w, providers=providers)
+        elif xml:  # OpenVINO
+            LOGGER.info(f'Loading {w} for OpenVINO inference...')
+            check_requirements(('openvino-dev',))  # requires openvino-dev: https://pypi.org/project/openvino-dev/
+            import openvino.inference_engine as ie
+            core = ie.IECore()
+            if not Path(w).is_file():  # if not *.xml
+                w = next(Path(w).glob('*.xml'))  # get *.xml file from *_openvino_model dir
+            network = core.read_network(model=w, weights=Path(w).with_suffix('.bin'))  # *.xml, *.bin paths
+            executable_network = core.load_network(network, device_name='CPU', num_requests=1)
+        elif engine:  # TensorRT
+            LOGGER.info(f'Loading {w} for TensorRT inference...')
+            import tensorrt as trt  # https://developer.nvidia.com/nvidia-tensorrt-download
+            check_version(trt.__version__, '7.0.0', hard=True)  # require tensorrt>=7.0.0
+            Binding = namedtuple('Binding', ('name', 'dtype', 'shape', 'data', 'ptr'))
+            logger = trt.Logger(trt.Logger.INFO)
+            with open(w, 'rb') as f, trt.Runtime(logger) as runtime:
+                model = runtime.deserialize_cuda_engine(f.read())
+            bindings = OrderedDict()
+            for index in range(model.num_bindings):
+                name = model.get_binding_name(index)
+                dtype = trt.nptype(model.get_binding_dtype(index))
+                shape = tuple(model.get_binding_shape(index))
+                data = torch.from_numpy(np.empty(shape, dtype=np.dtype(dtype))).to(device)
+                bindings[name] = Binding(name, dtype, shape, data, int(data.data_ptr()))
+            binding_addrs = OrderedDict((n, d.ptr) for n, d in bindings.items())
+            context = model.create_execution_context()
+            batch_size = bindings['images'].shape[0]
+        elif coreml:  # CoreML
+            LOGGER.info(f'Loading {w} for CoreML inference...')
+            import coremltools as ct
+            model = ct.models.MLModel(w)
+        else:  # TensorFlow (SavedModel, GraphDef, Lite, Edge TPU)
+            if saved_model:  # SavedModel
+                LOGGER.info(f'Loading {w} for TensorFlow SavedModel inference...')
+                import tensorflow as tf
+                keras = False  # assume TF1 saved_model
+                model = tf.keras.models.load_model(w) if keras else tf.saved_model.load(w)
+            elif pb:  # GraphDef https://www.tensorflow.org/guide/migrate#a_graphpb_or_graphpbtxt
+                LOGGER.info(f'Loading {w} for TensorFlow GraphDef inference...')
+                import tensorflow as tf
+
+                def wrap_frozen_graph(gd, inputs, outputs):
+                    x = tf.compat.v1.wrap_function(lambda: tf.compat.v1.import_graph_def(gd, name=""), [])  # wrapped
+                    ge = x.graph.as_graph_element
+                    return x.prune(tf.nest.map_structure(ge, inputs), tf.nest.map_structure(ge, outputs))
+
+                gd = tf.Graph().as_graph_def()  # graph_def
+                gd.ParseFromString(open(w, 'rb').read())
+                frozen_func = wrap_frozen_graph(gd, inputs="x:0", outputs="Identity:0")
+            elif tflite or edgetpu:  # https://www.tensorflow.org/lite/guide/python#install_tensorflow_lite_for_python
+                try:  # https://coral.ai/docs/edgetpu/tflite-python/#update-existing-tf-lite-code-for-the-edge-tpu
+                    from tflite_runtime.interpreter import Interpreter, load_delegate
+                except ImportError:
+                    import tensorflow as tf
+                    Interpreter, load_delegate = tf.lite.Interpreter, tf.lite.experimental.load_delegate,
+                if edgetpu:  # Edge TPU https://coral.ai/software/#edgetpu-runtime
+                    LOGGER.info(f'Loading {w} for TensorFlow Lite Edge TPU inference...')
+                    delegate = {'Linux': 'libedgetpu.so.1',
+                                'Darwin': 'libedgetpu.1.dylib',
+                                'Windows': 'edgetpu.dll'}[platform.system()]
+                    interpreter = Interpreter(model_path=w, experimental_delegates=[load_delegate(delegate)])
+                else:  # Lite
+                    LOGGER.info(f'Loading {w} for TensorFlow Lite inference...')
+                    interpreter = Interpreter(model_path=w)  # load TFLite model
+                interpreter.allocate_tensors()  # allocate
+                input_details = interpreter.get_input_details()  # inputs
+                output_details = interpreter.get_output_details()  # outputs
+            elif tfjs:
+                raise Exception('ERROR: YOLOv5 TF.js inference is not supported')
+        self.__dict__.update(locals())  # assign all variables to self
+
+    def forward(self, im, augment=False, visualize=False, val=False):
+        # YOLOv5 MultiBackend inference
+        b, ch, h, w = im.shape  # batch, channel, height, width
+        if self.pt or self.jit:  # PyTorch
+            y = self.model(im) if self.jit else self.model(im, augment=augment, visualize=visualize)
+            return y if val else y[0]
+        elif self.dnn:  # ONNX OpenCV DNN
+            im = im.cpu().numpy()  # torch to numpy
+            self.net.setInput(im)
+            y = self.net.forward()
+        elif self.onnx:  # ONNX Runtime
+            im = im.cpu().numpy()  # torch to numpy
+            y = self.session.run([self.session.get_outputs()[0].name], {self.session.get_inputs()[0].name: im})[0]
+        elif self.xml:  # OpenVINO
+            im = im.cpu().numpy()  # FP32
+            desc = self.ie.TensorDesc(precision='FP32', dims=im.shape, layout='NCHW')  # Tensor Description
+            request = self.executable_network.requests[0]  # inference request
+            request.set_blob(blob_name='images', blob=self.ie.Blob(desc, im))  # name=next(iter(request.input_blobs))
+            request.infer()
+            y = request.output_blobs['output'].buffer  # name=next(iter(request.output_blobs))
+        elif self.engine:  # TensorRT
+            assert im.shape == self.bindings['images'].shape, (im.shape, self.bindings['images'].shape)
+            self.binding_addrs['images'] = int(im.data_ptr())
+            self.context.execute_v2(list(self.binding_addrs.values()))
+            y = self.bindings['output'].data
+        elif self.coreml:  # CoreML
+            im = im.permute(0, 2, 3, 1).cpu().numpy()  # torch BCHW to numpy BHWC shape(1,320,192,3)
+            im = Image.fromarray((im[0] * 255).astype('uint8'))
+            # im = im.resize((192, 320), Image.ANTIALIAS)
+            y = self.model.predict({'image': im})  # coordinates are xywh normalized
+            if 'confidence' in y:
+                box = xywh2xyxy(y['coordinates'] * [[w, h, w, h]])  # xyxy pixels
+                conf, cls = y['confidence'].max(1), y['confidence'].argmax(1).astype(np.float)
+                y = np.concatenate((box, conf.reshape(-1, 1), cls.reshape(-1, 1)), 1)
+            else:
+                k = 'var_' + str(sorted(int(k.replace('var_', '')) for k in y)[-1])  # output key
+                y = y[k]  # output
+        else:  # TensorFlow (SavedModel, GraphDef, Lite, Edge TPU)
+            im = im.permute(0, 2, 3, 1).cpu().numpy()  # torch BCHW to numpy BHWC shape(1,320,192,3)
+            if self.saved_model:  # SavedModel
+                y = (self.model(im, training=False) if self.keras else self.model(im)[0]).numpy()
+            elif self.pb:  # GraphDef
+                y = self.frozen_func(x=self.tf.constant(im)).numpy()
+            else:  # Lite or Edge TPU
+                input, output = self.input_details[0], self.output_details[0]
+                int8 = input['dtype'] == np.uint8  # is TFLite quantized uint8 model
+                if int8:
+                    scale, zero_point = input['quantization']
+                    im = (im / scale + zero_point).astype(np.uint8)  # de-scale
+                self.interpreter.set_tensor(input['index'], im)
+                self.interpreter.invoke()
+                y = self.interpreter.get_tensor(output['index'])
+                if int8:
+                    scale, zero_point = output['quantization']
+                    y = (y.astype(np.float32) - zero_point) * scale  # re-scale
+            y[..., :4] *= [w, h, w, h]  # xywh normalized to pixels
+
+        y = torch.tensor(y) if isinstance(y, np.ndarray) else y
+        return (y, []) if val else y
+
+    def warmup(self, imgsz=(1, 3, 640, 640), half=False):
+        # Warmup model by running inference once
+        if self.pt or self.jit or self.onnx or self.engine:  # warmup types
+            if isinstance(self.device, torch.device) and self.device.type != 'cpu':  # only warmup GPU models
+                im = torch.zeros(*imgsz).to(self.device).type(torch.half if half else torch.float)  # input image
+                self.forward(im)  # warmup
+
+    @staticmethod
+    def model_type(p='path/to/model.pt'):
+        # Return model type from model path, i.e. path='path/to/model.onnx' -> type=onnx
+        from export import export_formats
+        suffixes = list(export_formats().Suffix) + ['.xml']  # export suffixes
+        check_suffix(p, suffixes)  # checks
+        p = Path(p).name  # eliminate trailing separators
+        pt, jit, onnx, xml, engine, coreml, saved_model, pb, tflite, edgetpu, tfjs, xml2 = (s in p for s in suffixes)
+        xml |= xml2  # *_openvino_model or *.xml
+        tflite &= not edgetpu  # *.tflite
+        return pt, jit, onnx, xml, engine, coreml, saved_model, pb, tflite, edgetpu, tfjs
+
+class DetectPrunedMultiBackend(nn.Module):
+    # YOLOv5 MultiBackend class for python inference on various backends
+    def __init__(self, weights='yolov5s.pt', device=None, dnn=False, data=None):
+        # Usage:
+        #   PyTorch:              weights = *.pt
+        #   TorchScript:                    *.torchscript
+        #   ONNX Runtime:                   *.onnx
+        #   ONNX OpenCV DNN:                *.onnx with --dnn
+        #   OpenVINO:                       *.xml
+        #   CoreML:                         *.mlmodel
+        #   TensorRT:                       *.engine
+        from models.experimental import attempt_download, attempt_load_pruned  # scoped to avoid circular import
+
+        super().__init__()
+        w = str(weights[0] if isinstance(weights, list) else weights)
+        pt, jit, onnx, xml, engine, coreml, saved_model, pb, tflite, edgetpu, tfjs = self.model_type(w)  # get backend
+        stride, names = 64, [f'class{i}' for i in range(1000)]  # assign defaults
+        w = attempt_download(w)  # download if not local
+        if data:  # data.yaml path (optional)
+            with open(data, errors='ignore') as f:
+                names = yaml.safe_load(f)['names']  # class names
+
+        if pt:  # PyTorch
+            model = attempt_load_pruned(weights if isinstance(weights, list) else w, map_location=device)
+            stride = max(int(model.stride.max()), 32)  # model stride
+            names = model.module.names if hasattr(model, 'module') else model.names  # get class names
+            self.model = model  # explicitly assign for to(), cpu(), cuda(), half()
+        elif jit:  # TorchScript
+            LOGGER.info(f'Loading {w} for TorchScript inference...')
+            extra_files = {'config.txt': ''}  # model metadata
+            model = torch.jit.load(w, _extra_files=extra_files)
+            if extra_files['config.txt']:
+                d = json.loads(extra_files['config.txt'])  # extra_files dict
+                stride, names = int(d['stride']), d['names']
+        elif dnn:  # ONNX OpenCV DNN
+            LOGGER.info(f'Loading {w} for ONNX OpenCV DNN inference...')
+            check_requirements(('opencv-python>=4.5.4',))
+            net = cv2.dnn.readNetFromONNX(w)
+        elif onnx:  # ONNX Runtime
+            LOGGER.info(f'Loading {w} for ONNX Runtime inference...')
+            cuda = torch.cuda.is_available()
+            check_requirements(('onnx', 'onnxruntime-gpu' if cuda else 'onnxruntime'))
+            import onnxruntime
+            providers = ['CUDAExecutionProvider', 'CPUExecutionProvider'] if cuda else ['CPUExecutionProvider']
+            session = onnxruntime.InferenceSession(w, providers=providers)
+        elif xml:  # OpenVINO
+            LOGGER.info(f'Loading {w} for OpenVINO inference...')
+            check_requirements(('openvino-dev',))  # requires openvino-dev: https://pypi.org/project/openvino-dev/
+            import openvino.inference_engine as ie
+            core = ie.IECore()
+            if not Path(w).is_file():  # if not *.xml
+                w = next(Path(w).glob('*.xml'))  # get *.xml file from *_openvino_model dir
+            network = core.read_network(model=w, weights=Path(w).with_suffix('.bin'))  # *.xml, *.bin paths
+            executable_network = core.load_network(network, device_name='CPU', num_requests=1)
+        elif engine:  # TensorRT
+            LOGGER.info(f'Loading {w} for TensorRT inference...')
+            import tensorrt as trt  # https://developer.nvidia.com/nvidia-tensorrt-download
+            check_version(trt.__version__, '7.0.0', hard=True)  # require tensorrt>=7.0.0
+            Binding = namedtuple('Binding', ('name', 'dtype', 'shape', 'data', 'ptr'))
+            logger = trt.Logger(trt.Logger.INFO)
+            with open(w, 'rb') as f, trt.Runtime(logger) as runtime:
+                model = runtime.deserialize_cuda_engine(f.read())
+            bindings = OrderedDict()
+            for index in range(model.num_bindings):
+                name = model.get_binding_name(index)
+                dtype = trt.nptype(model.get_binding_dtype(index))
+                shape = tuple(model.get_binding_shape(index))
+                data = torch.from_numpy(np.empty(shape, dtype=np.dtype(dtype))).to(device)
+                bindings[name] = Binding(name, dtype, shape, data, int(data.data_ptr()))
+            binding_addrs = OrderedDict((n, d.ptr) for n, d in bindings.items())
+            context = model.create_execution_context()
+            batch_size = bindings['images'].shape[0]
+        elif coreml:  # CoreML
+            LOGGER.info(f'Loading {w} for CoreML inference...')
+            import coremltools as ct
+            model = ct.models.MLModel(w)
+        else:  # TensorFlow (SavedModel, GraphDef, Lite, Edge TPU)
+            if saved_model:  # SavedModel
+                LOGGER.info(f'Loading {w} for TensorFlow SavedModel inference...')
+                import tensorflow as tf
+                keras = False  # assume TF1 saved_model
+                model = tf.keras.models.load_model(w) if keras else tf.saved_model.load(w)
+            elif pb:  # GraphDef https://www.tensorflow.org/guide/migrate#a_graphpb_or_graphpbtxt
+                LOGGER.info(f'Loading {w} for TensorFlow GraphDef inference...')
+                import tensorflow as tf
+
+                def wrap_frozen_graph(gd, inputs, outputs):
+                    x = tf.compat.v1.wrap_function(lambda: tf.compat.v1.import_graph_def(gd, name=""), [])  # wrapped
+                    ge = x.graph.as_graph_element
+                    return x.prune(tf.nest.map_structure(ge, inputs), tf.nest.map_structure(ge, outputs))
+
+                gd = tf.Graph().as_graph_def()  # graph_def
+                gd.ParseFromString(open(w, 'rb').read())
+                frozen_func = wrap_frozen_graph(gd, inputs="x:0", outputs="Identity:0")
+            elif tflite or edgetpu:  # https://www.tensorflow.org/lite/guide/python#install_tensorflow_lite_for_python
+                try:  # https://coral.ai/docs/edgetpu/tflite-python/#update-existing-tf-lite-code-for-the-edge-tpu
+                    from tflite_runtime.interpreter import Interpreter, load_delegate
+                except ImportError:
+                    import tensorflow as tf
+                    Interpreter, load_delegate = tf.lite.Interpreter, tf.lite.experimental.load_delegate,
+                if edgetpu:  # Edge TPU https://coral.ai/software/#edgetpu-runtime
+                    LOGGER.info(f'Loading {w} for TensorFlow Lite Edge TPU inference...')
+                    delegate = {'Linux': 'libedgetpu.so.1',
+                                'Darwin': 'libedgetpu.1.dylib',
+                                'Windows': 'edgetpu.dll'}[platform.system()]
+                    interpreter = Interpreter(model_path=w, experimental_delegates=[load_delegate(delegate)])
+                else:  # Lite
+                    LOGGER.info(f'Loading {w} for TensorFlow Lite inference...')
+                    interpreter = Interpreter(model_path=w)  # load TFLite model
+                interpreter.allocate_tensors()  # allocate
+                input_details = interpreter.get_input_details()  # inputs
+                output_details = interpreter.get_output_details()  # outputs
+            elif tfjs:
+                raise Exception('ERROR: YOLOv5 TF.js inference is not supported')
+        self.__dict__.update(locals())  # assign all variables to self
+
+    def forward(self, im, augment=False, visualize=False, val=False):
+        # YOLOv5 MultiBackend inference
+        b, ch, h, w = im.shape  # batch, channel, height, width
+        if self.pt or self.jit:  # PyTorch
+            y = self.model(im) if self.jit else self.model(im, augment=augment, visualize=visualize)
+            return y if val else y[0]
+        elif self.dnn:  # ONNX OpenCV DNN
+            im = im.cpu().numpy()  # torch to numpy
+            self.net.setInput(im)
+            y = self.net.forward()
+        elif self.onnx:  # ONNX Runtime
+            im = im.cpu().numpy()  # torch to numpy
+            y = self.session.run([self.session.get_outputs()[0].name], {self.session.get_inputs()[0].name: im})[0]
+        elif self.xml:  # OpenVINO
+            im = im.cpu().numpy()  # FP32
+            desc = self.ie.TensorDesc(precision='FP32', dims=im.shape, layout='NCHW')  # Tensor Description
+            request = self.executable_network.requests[0]  # inference request
+            request.set_blob(blob_name='images', blob=self.ie.Blob(desc, im))  # name=next(iter(request.input_blobs))
+            request.infer()
+            y = request.output_blobs['output'].buffer  # name=next(iter(request.output_blobs))
+        elif self.engine:  # TensorRT
+            assert im.shape == self.bindings['images'].shape, (im.shape, self.bindings['images'].shape)
+            self.binding_addrs['images'] = int(im.data_ptr())
+            self.context.execute_v2(list(self.binding_addrs.values()))
+            y = self.bindings['output'].data
+        elif self.coreml:  # CoreML
+            im = im.permute(0, 2, 3, 1).cpu().numpy()  # torch BCHW to numpy BHWC shape(1,320,192,3)
+            im = Image.fromarray((im[0] * 255).astype('uint8'))
+            # im = im.resize((192, 320), Image.ANTIALIAS)
+            y = self.model.predict({'image': im})  # coordinates are xywh normalized
+            if 'confidence' in y:
+                box = xywh2xyxy(y['coordinates'] * [[w, h, w, h]])  # xyxy pixels
+                conf, cls = y['confidence'].max(1), y['confidence'].argmax(1).astype(np.float)
+                y = np.concatenate((box, conf.reshape(-1, 1), cls.reshape(-1, 1)), 1)
+            else:
+                k = 'var_' + str(sorted(int(k.replace('var_', '')) for k in y)[-1])  # output key
+                y = y[k]  # output
+        else:  # TensorFlow (SavedModel, GraphDef, Lite, Edge TPU)
+            im = im.permute(0, 2, 3, 1).cpu().numpy()  # torch BCHW to numpy BHWC shape(1,320,192,3)
+            if self.saved_model:  # SavedModel
+                y = (self.model(im, training=False) if self.keras else self.model(im)[0]).numpy()
+            elif self.pb:  # GraphDef
+                y = self.frozen_func(x=self.tf.constant(im)).numpy()
+            else:  # Lite or Edge TPU
+                input, output = self.input_details[0], self.output_details[0]
+                int8 = input['dtype'] == np.uint8  # is TFLite quantized uint8 model
+                if int8:
+                    scale, zero_point = input['quantization']
+                    im = (im / scale + zero_point).astype(np.uint8)  # de-scale
+                self.interpreter.set_tensor(input['index'], im)
+                self.interpreter.invoke()
+                y = self.interpreter.get_tensor(output['index'])
+                if int8:
+                    scale, zero_point = output['quantization']
+                    y = (y.astype(np.float32) - zero_point) * scale  # re-scale
+            y[..., :4] *= [w, h, w, h]  # xywh normalized to pixels
+
+        y = torch.tensor(y) if isinstance(y, np.ndarray) else y
+        return (y, []) if val else y
+
+    def warmup(self, imgsz=(1, 3, 640, 640), half=False):
+        # Warmup model by running inference once
+        if self.pt or self.jit or self.onnx or self.engine:  # warmup types
+            if isinstance(self.device, torch.device) and self.device.type != 'cpu':  # only warmup GPU models
+                im = torch.zeros(*imgsz).to(self.device).type(torch.half if half else torch.float)  # input image
+                self.forward(im)  # warmup
+
+    @staticmethod
+    def model_type(p='path/to/model.pt'):
+        # Return model type from model path, i.e. path='path/to/model.onnx' -> type=onnx
+        from export import export_formats
+        suffixes = list(export_formats().Suffix) + ['.xml']  # export suffixes
+        check_suffix(p, suffixes)  # checks
+        p = Path(p).name  # eliminate trailing separators
+        pt, jit, onnx, xml, engine, coreml, saved_model, pb, tflite, edgetpu, tfjs, xml2 = (s in p for s in suffixes)
+        xml |= xml2  # *_openvino_model or *.xml
+        tflite &= not edgetpu  # *.tflite
+        return pt, jit, onnx, xml, engine, coreml, saved_model, pb, tflite, edgetpu, tfjs
+        
+class AutoShape(nn.Module):
+    # YOLOv5 input-robust model wrapper for passing cv2/np/PIL/torch inputs. Includes preprocessing, inference and NMS
+    conf = 0.25  # NMS confidence threshold
+    iou = 0.45  # NMS IoU threshold
+    agnostic = False  # NMS class-agnostic
+    multi_label = False  # NMS multiple labels per box
+    classes = None  # (optional list) filter by class, i.e. = [0, 15, 16] for COCO persons, cats and dogs
+    max_det = 1000  # maximum number of detections per image
+    amp = False  # Automatic Mixed Precision (AMP) inference
+
+    def __init__(self, model):
+        super().__init__()
+        LOGGER.info('Adding AutoShape... ')
+        copy_attr(self, model, include=('yaml', 'nc', 'hyp', 'names', 'stride', 'abc'), exclude=())  # copy attributes
+        self.dmb = isinstance(model, DetectMultiBackend)  # DetectMultiBackend() instance
+        self.pt = not self.dmb or model.pt  # PyTorch model
+        self.model = model.eval()
+
+    def _apply(self, fn):
+        # Apply to(), cpu(), cuda(), half() to model tensors that are not parameters or registered buffers
+        self = super()._apply(fn)
+        if self.pt:
+            m = self.model.model.model[-1] if self.dmb else self.model.model[-1]  # Detect()
+            m.stride = fn(m.stride)
+            m.grid = list(map(fn, m.grid))
+            if isinstance(m.anchor_grid, list):
+                m.anchor_grid = list(map(fn, m.anchor_grid))
+        return self
+
+    @torch.no_grad()
+    def forward(self, imgs, size=640, augment=False, profile=False):
+        # Inference from various sources. For height=640, width=1280, RGB images example inputs are:
+        #   file:       imgs = 'data/images/zidane.jpg'  # str or PosixPath
+        #   URI:             = 'https://ultralytics.com/images/zidane.jpg'
+        #   OpenCV:          = cv2.imread('image.jpg')[:,:,::-1]  # HWC BGR to RGB x(640,1280,3)
+        #   PIL:             = Image.open('image.jpg') or ImageGrab.grab()  # HWC x(640,1280,3)
+        #   numpy:           = np.zeros((640,1280,3))  # HWC
+        #   torch:           = torch.zeros(16,3,320,640)  # BCHW (scaled to size=640, 0-1 values)
+        #   multiple:        = [Image.open('image1.jpg'), Image.open('image2.jpg'), ...]  # list of images
+
+        t = [time_sync()]
+        p = next(self.model.parameters()) if self.pt else torch.zeros(1)  # for device and type
+        autocast = self.amp and (p.device.type != 'cpu')  # Automatic Mixed Precision (AMP) inference
+        if isinstance(imgs, torch.Tensor):  # torch
+            with amp.autocast(enabled=autocast):
+                return self.model(imgs.to(p.device).type_as(p), augment, profile)  # inference
+
+        # Pre-process
+        n, imgs = (len(imgs), imgs) if isinstance(imgs, list) else (1, [imgs])  # number of images, list of images
+        shape0, shape1, files = [], [], []  # image and inference shapes, filenames
+        for i, im in enumerate(imgs):
+            f = f'image{i}'  # filename
+            if isinstance(im, (str, Path)):  # filename or uri
+                im, f = Image.open(requests.get(im, stream=True).raw if str(im).startswith('http') else im), im
+                im = np.asarray(exif_transpose(im))
+            elif isinstance(im, Image.Image):  # PIL Image
+                im, f = np.asarray(exif_transpose(im)), getattr(im, 'filename', f) or f
+            files.append(Path(f).with_suffix('.jpg').name)
+            if im.shape[0] < 5:  # image in CHW
+                im = im.transpose((1, 2, 0))  # reverse dataloader .transpose(2, 0, 1)
+            im = im[..., :3] if im.ndim == 3 else np.tile(im[..., None], 3)  # enforce 3ch input
+            s = im.shape[:2]  # HWC
+            shape0.append(s)  # image shape
+            g = (size / max(s))  # gain
+            shape1.append([y * g for y in s])
+            imgs[i] = im if im.data.contiguous else np.ascontiguousarray(im)  # update
+        shape1 = [make_divisible(x, self.stride) for x in np.stack(shape1, 0).max(0)]  # inference shape
+        x = [letterbox(im, new_shape=shape1 if self.pt else size, auto=False)[0] for im in imgs]  # pad
+        x = np.stack(x, 0) if n > 1 else x[0][None]  # stack
+        x = np.ascontiguousarray(x.transpose((0, 3, 1, 2)))  # BHWC to BCHW
+        x = torch.from_numpy(x).to(p.device).type_as(p) / 255  # uint8 to fp16/32
+        t.append(time_sync())
+
+        with amp.autocast(enabled=autocast):
+            # Inference
+            y = self.model(x, augment, profile)  # forward
+            t.append(time_sync())
+
+            # Post-process
+            y = non_max_suppression(y if self.dmb else y[0], self.conf, iou_thres=self.iou, classes=self.classes,
+                                    agnostic=self.agnostic, multi_label=self.multi_label, max_det=self.max_det)  # NMS
+            for i in range(n):
+                scale_coords(shape1, y[i][:, :4], shape0[i])
+
+            t.append(time_sync())
+            return Detections(imgs, y, files, t, self.names, x.shape)
+
+
+class Detections:
+    # YOLOv5 detections class for inference results
+    def __init__(self, imgs, pred, files, times=(0, 0, 0, 0), names=None, shape=None):
+        super().__init__()
+        d = pred[0].device  # device
+        gn = [torch.tensor([*(im.shape[i] for i in [1, 0, 1, 0]), 1, 1], device=d) for im in imgs]  # normalizations
+        self.imgs = imgs  # list of images as numpy arrays
+        self.pred = pred  # list of tensors pred[0] = (xyxy, conf, cls)
+        self.names = names  # class names
+        self.files = files  # image filenames
+        self.times = times  # profiling times
+        self.xyxy = pred  # xyxy pixels
+        self.xywh = [xyxy2xywh(x) for x in pred]  # xywh pixels
+        self.xyxyn = [x / g for x, g in zip(self.xyxy, gn)]  # xyxy normalized
+        self.xywhn = [x / g for x, g in zip(self.xywh, gn)]  # xywh normalized
+        self.n = len(self.pred)  # number of images (batch size)
+        self.t = tuple((times[i + 1] - times[i]) * 1000 / self.n for i in range(3))  # timestamps (ms)
+        self.s = shape  # inference BCHW shape
+
+    def display(self, pprint=False, show=False, save=False, crop=False, render=False, save_dir=Path('')):
+        crops = []
+        for i, (im, pred) in enumerate(zip(self.imgs, self.pred)):
+            s = f'image {i + 1}/{len(self.pred)}: {im.shape[0]}x{im.shape[1]} '  # string
+            if pred.shape[0]:
+                for c in pred[:, -1].unique():
+                    n = (pred[:, -1] == c).sum()  # detections per class
+                    s += f"{n} {self.names[int(c)]}{'s' * (n > 1)}, "  # add to string
+                if show or save or render or crop:
+                    annotator = Annotator(im, example=str(self.names))
+                    for *box, conf, cls in reversed(pred):  # xyxy, confidence, class
+                        label = f'{self.names[int(cls)]} {conf:.2f}'
+                        if crop:
+                            file = save_dir / 'crops' / self.names[int(cls)] / self.files[i] if save else None
+                            crops.append({'box': box, 'conf': conf, 'cls': cls, 'label': label,
+                                          'im': save_one_box(box, im, file=file, save=save)})
+                        else:  # all others
+                            annotator.box_label(box, label, color=colors(cls))
+                    im = annotator.im
+            else:
+                s += '(no detections)'
+
+            im = Image.fromarray(im.astype(np.uint8)) if isinstance(im, np.ndarray) else im  # from np
+            if pprint:
+                LOGGER.info(s.rstrip(', '))
+            if show:
+                im.show(self.files[i])  # show
+            if save:
+                f = self.files[i]
+                im.save(save_dir / f)  # save
+                if i == self.n - 1:
+                    LOGGER.info(f"Saved {self.n} image{'s' * (self.n > 1)} to {colorstr('bold', save_dir)}")
+            if render:
+                self.imgs[i] = np.asarray(im)
+        if crop:
+            if save:
+                LOGGER.info(f'Saved results to {save_dir}\n')
+            return crops
+
+    def print(self):
+        self.display(pprint=True)  # print results
+        LOGGER.info(f'Speed: %.1fms pre-process, %.1fms inference, %.1fms NMS per image at shape {tuple(self.s)}' %
+                    self.t)
+
+    def show(self):
+        self.display(show=True)  # show results
+
+    def save(self, save_dir='runs/detect/exp'):
+        save_dir = increment_path(save_dir, exist_ok=save_dir != 'runs/detect/exp', mkdir=True)  # increment save_dir
+        self.display(save=True, save_dir=save_dir)  # save results
+
+    def crop(self, save=True, save_dir='runs/detect/exp'):
+        save_dir = increment_path(save_dir, exist_ok=save_dir != 'runs/detect/exp', mkdir=True) if save else None
+        return self.display(crop=True, save=save, save_dir=save_dir)  # crop results
+
+    def render(self):
+        self.display(render=True)  # render results
+        return self.imgs
+
+    def pandas(self):
+        # return detections as pandas DataFrames, i.e. print(results.pandas().xyxy[0])
+        new = copy(self)  # return copy
+        ca = 'xmin', 'ymin', 'xmax', 'ymax', 'confidence', 'class', 'name'  # xyxy columns
+        cb = 'xcenter', 'ycenter', 'width', 'height', 'confidence', 'class', 'name'  # xywh columns
+        for k, c in zip(['xyxy', 'xyxyn', 'xywh', 'xywhn'], [ca, ca, cb, cb]):
+            a = [[x[:5] + [int(x[5]), self.names[int(x[5])]] for x in x.tolist()] for x in getattr(self, k)]  # update
+            setattr(new, k, [pd.DataFrame(x, columns=c) for x in a])
+        return new
+
+    def tolist(self):
+        # return a list of Detections objects, i.e. 'for result in results.tolist():'
+        r = range(self.n)  # iterable
+        x = [Detections([self.imgs[i]], [self.pred[i]], [self.files[i]], self.times, self.names, self.s) for i in r]
+        # for d in x:
+        #    for k in ['imgs', 'pred', 'xyxy', 'xyxyn', 'xywh', 'xywhn']:
+        #        setattr(d, k, getattr(d, k)[0])  # pop out of list
+        return x
+
+    def __len__(self):
+        return self.n
+
+
+class Classify(nn.Module):
+    # Classification head, i.e. x(b,c1,20,20) to x(b,c2)
+    def __init__(self, c1, c2, k=1, s=1, p=None, g=1):  # ch_in, ch_out, kernel, stride, padding, groups
+        super().__init__()
+        self.aap = nn.AdaptiveAvgPool2d(1)  # to x(b,c1,1,1)
+        self.conv = nn.Conv2d(c1, c2, k, s, autopad(k, p), groups=g)  # to x(b,c2,1,1)
+        self.flat = nn.Flatten()
+
+    def forward(self, x):
+        z = torch.cat([self.aap(y) for y in (x if isinstance(x, list) else [x])], 1)  # cat if list
+        return self.flat(self.conv(z))  # flatten to x(b,c2)

models/experimental.py

@@ -0,0 +1,155 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+"""
+Experimental modules
+"""
+import math
+
+import numpy as np
+import torch
+import torch.nn as nn
+
+from models.common import Conv
+from utils.downloads import attempt_download
+
+
+class CrossConv(nn.Module):
+    # Cross Convolution Downsample
+    def __init__(self, c1, c2, k=3, s=1, g=1, e=1.0, shortcut=False):
+        # ch_in, ch_out, kernel, stride, groups, expansion, shortcut
+        super().__init__()
+        c_ = int(c2 * e)  # hidden channels
+        self.cv1 = Conv(c1, c_, (1, k), (1, s))
+        self.cv2 = Conv(c_, c2, (k, 1), (s, 1), g=g)
+        self.add = shortcut and c1 == c2
+
+    def forward(self, x):
+        return x + self.cv2(self.cv1(x)) if self.add else self.cv2(self.cv1(x))
+
+
+class Sum(nn.Module):
+    # Weighted sum of 2 or more layers https://arxiv.org/abs/1911.09070
+    def __init__(self, n, weight=False):  # n: number of inputs
+        super().__init__()
+        self.weight = weight  # apply weights boolean
+        self.iter = range(n - 1)  # iter object
+        if weight:
+            self.w = nn.Parameter(-torch.arange(1.0, n) / 2, requires_grad=True)  # layer weights
+
+    def forward(self, x):
+        y = x[0]  # no weight
+        if self.weight:
+            w = torch.sigmoid(self.w) * 2
+            for i in self.iter:
+                y = y + x[i + 1] * w[i]
+        else:
+            for i in self.iter:
+                y = y + x[i + 1]
+        return y
+
+
+class MixConv2d(nn.Module):
+    # Mixed Depth-wise Conv https://arxiv.org/abs/1907.09595
+    def __init__(self, c1, c2, k=(1, 3), s=1, equal_ch=True):  # ch_in, ch_out, kernel, stride, ch_strategy
+        super().__init__()
+        n = len(k)  # number of convolutions
+        if equal_ch:  # equal c_ per group
+            i = torch.linspace(0, n - 1E-6, c2).floor()  # c2 indices
+            c_ = [(i == g).sum() for g in range(n)]  # intermediate channels
+        else:  # equal weight.numel() per group
+            b = [c2] + [0] * n
+            a = np.eye(n + 1, n, k=-1)
+            a -= np.roll(a, 1, axis=1)
+            a *= np.array(k) ** 2
+            a[0] = 1
+            c_ = np.linalg.lstsq(a, b, rcond=None)[0].round()  # solve for equal weight indices, ax = b
+
+        self.m = nn.ModuleList(
+            [nn.Conv2d(c1, int(c_), k, s, k // 2, groups=math.gcd(c1, int(c_)), bias=False) for k, c_ in zip(k, c_)])
+        self.bn = nn.BatchNorm2d(c2)
+        self.act = nn.SiLU()
+
+    def forward(self, x):
+        return self.act(self.bn(torch.cat([m(x) for m in self.m], 1)))
+
+
+class Ensemble(nn.ModuleList):
+    # Ensemble of models
+    def __init__(self):
+        super().__init__()
+
+    def forward(self, x, augment=False, profile=False, visualize=False):
+        y = []
+        for module in self:
+            y.append(module(x, augment, profile, visualize)[0])
+        # y = torch.stack(y).max(0)[0]  # max ensemble
+        # y = torch.stack(y).mean(0)  # mean ensemble
+        y = torch.cat(y, 1)  # nms ensemble
+        return y, None  # inference, train output
+
+
+def attempt_load(weights, map_location=None, inplace=True, fuse=True):
+    from models.yolo import Detect, Model
+
+    # Loads an ensemble of models weights=[a,b,c] or a single model weights=[a] or weights=a
+    model = Ensemble()
+    for w in weights if isinstance(weights, list) else [weights]:
+        ckpt = torch.load(attempt_download(w), map_location=map_location)  # load
+        if fuse:
+            # model.append(ckpt['ema' if ckpt.get('ema') else 'model'].float().fuse().eval())  # FP32 model
+            model.append(ckpt['ema' if ckpt.get('ema') else 'model'].float().eval())  # no fuse for bn prune
+        else:
+            model.append(ckpt['ema' if ckpt.get('ema') else 'model'].float().eval())  # without layer fuse
+
+    # Compatibility updates
+    for m in model.modules():
+        if type(m) in [nn.Hardswish, nn.LeakyReLU, nn.ReLU, nn.ReLU6, nn.SiLU, Detect, Model]:
+            m.inplace = inplace  # pytorch 1.7.0 compatibility
+            if type(m) is Detect:
+                if not isinstance(m.anchor_grid, list):  # new Detect Layer compatibility
+                    delattr(m, 'anchor_grid')
+                    setattr(m, 'anchor_grid', [torch.zeros(1)] * m.nl)
+        elif type(m) is Conv:
+            m._non_persistent_buffers_set = set()  # pytorch 1.6.0 compatibility
+
+    if len(model) == 1:
+        return model[-1]  # return model
+    else:
+        print(f'Ensemble created with {weights}\n')
+        for k in ['names']:
+            setattr(model, k, getattr(model[-1], k))
+        model.stride = model[torch.argmax(torch.tensor([m.stride.max() for m in model])).int()].stride  # max stride
+        return model  # return ensemble
+
+
+def attempt_load_pruned(weights, map_location=None, inplace=True, fuse=True):
+    from models.yolo import Detect, ModelPruned
+
+    # Loads an ensemble of models weights=[a,b,c] or a single model weights=[a] or weights=a
+    model = Ensemble()
+    for w in weights if isinstance(weights, list) else [weights]:
+        ckpt = torch.load(attempt_download(w), map_location=map_location)  # load
+        if fuse:
+            # model.append(ckpt['ema' if ckpt.get('ema') else 'model'].float().fuse().eval())  # FP32 model
+            model.append(ckpt['ema' if ckpt.get('ema') else 'model'].float().eval())  # no fuse for bn prune
+        else:
+            model.append(ckpt['ema' if ckpt.get('ema') else 'model'].float().eval())  # without layer fuse
+
+    # Compatibility updates
+    for m in model.modules():
+        if type(m) in [nn.Hardswish, nn.LeakyReLU, nn.ReLU, nn.ReLU6, nn.SiLU, Detect, ModelPruned]:
+            m.inplace = inplace  # pytorch 1.7.0 compatibility
+            if type(m) is Detect:
+                if not isinstance(m.anchor_grid, list):  # new Detect Layer compatibility
+                    delattr(m, 'anchor_grid')
+                    setattr(m, 'anchor_grid', [torch.zeros(1)] * m.nl)
+        elif type(m) is Conv:
+            m._non_persistent_buffers_set = set()  # pytorch 1.6.0 compatibility
+
+    if len(model) == 1:
+        return model[-1]  # return model
+    else:
+        print(f'Ensemble created with {weights}\n')
+        for k in ['names']:
+            setattr(model, k, getattr(model[-1], k))
+        model.stride = model[torch.argmax(torch.tensor([m.stride.max() for m in model])).int()].stride  # max stride
+        return model  # return ensemble

models/hub/anchors.yaml

@@ -0,0 +1,59 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+# Default anchors for COCO data
+
+
+# P5 -------------------------------------------------------------------------------------------------------------------
+# P5-640:
+anchors_p5_640:
+  - [10,13, 16,30, 33,23]  # P3/8
+  - [30,61, 62,45, 59,119]  # P4/16
+  - [116,90, 156,198, 373,326]  # P5/32
+
+
+# P6 -------------------------------------------------------------------------------------------------------------------
+# P6-640:  thr=0.25: 0.9964 BPR, 5.54 anchors past thr, n=12, img_size=640, metric_all=0.281/0.716-mean/best, past_thr=0.469-mean: 9,11,  21,19,  17,41,  43,32,  39,70,  86,64,  65,131,  134,130,  120,265,  282,180,  247,354,  512,387
+anchors_p6_640:
+  - [9,11,  21,19,  17,41]  # P3/8
+  - [43,32,  39,70,  86,64]  # P4/16
+  - [65,131,  134,130,  120,265]  # P5/32
+  - [282,180,  247,354,  512,387]  # P6/64
+
+# P6-1280:  thr=0.25: 0.9950 BPR, 5.55 anchors past thr, n=12, img_size=1280, metric_all=0.281/0.714-mean/best, past_thr=0.468-mean: 19,27,  44,40,  38,94,  96,68,  86,152,  180,137,  140,301,  303,264,  238,542,  436,615,  739,380,  925,792
+anchors_p6_1280:
+  - [19,27,  44,40,  38,94]  # P3/8
+  - [96,68,  86,152,  180,137]  # P4/16
+  - [140,301,  303,264,  238,542]  # P5/32
+  - [436,615,  739,380,  925,792]  # P6/64
+
+# P6-1920:  thr=0.25: 0.9950 BPR, 5.55 anchors past thr, n=12, img_size=1920, metric_all=0.281/0.714-mean/best, past_thr=0.468-mean: 28,41,  67,59,  57,141,  144,103,  129,227,  270,205,  209,452,  455,396,  358,812,  653,922,  1109,570,  1387,1187
+anchors_p6_1920:
+  - [28,41,  67,59,  57,141]  # P3/8
+  - [144,103,  129,227,  270,205]  # P4/16
+  - [209,452,  455,396,  358,812]  # P5/32
+  - [653,922,  1109,570,  1387,1187]  # P6/64
+
+
+# P7 -------------------------------------------------------------------------------------------------------------------
+# P7-640:  thr=0.25: 0.9962 BPR, 6.76 anchors past thr, n=15, img_size=640, metric_all=0.275/0.733-mean/best, past_thr=0.466-mean: 11,11,  13,30,  29,20,  30,46,  61,38,  39,92,  78,80,  146,66,  79,163,  149,150,  321,143,  157,303,  257,402,  359,290,  524,372
+anchors_p7_640:
+  - [11,11,  13,30,  29,20]  # P3/8
+  - [30,46,  61,38,  39,92]  # P4/16
+  - [78,80,  146,66,  79,163]  # P5/32
+  - [149,150,  321,143,  157,303]  # P6/64
+  - [257,402,  359,290,  524,372]  # P7/128
+
+# P7-1280:  thr=0.25: 0.9968 BPR, 6.71 anchors past thr, n=15, img_size=1280, metric_all=0.273/0.732-mean/best, past_thr=0.463-mean: 19,22,  54,36,  32,77,  70,83,  138,71,  75,173,  165,159,  148,334,  375,151,  334,317,  251,626,  499,474,  750,326,  534,814,  1079,818
+anchors_p7_1280:
+  - [19,22,  54,36,  32,77]  # P3/8
+  - [70,83,  138,71,  75,173]  # P4/16
+  - [165,159,  148,334,  375,151]  # P5/32
+  - [334,317,  251,626,  499,474]  # P6/64
+  - [750,326,  534,814,  1079,818]  # P7/128
+
+# P7-1920:  thr=0.25: 0.9968 BPR, 6.71 anchors past thr, n=15, img_size=1920, metric_all=0.273/0.732-mean/best, past_thr=0.463-mean: 29,34,  81,55,  47,115,  105,124,  207,107,  113,259,  247,238,  222,500,  563,227,  501,476,  376,939,  749,711,  1126,489,  801,1222,  1618,1227
+anchors_p7_1920:
+  - [29,34,  81,55,  47,115]  # P3/8
+  - [105,124,  207,107,  113,259]  # P4/16
+  - [247,238,  222,500,  563,227]  # P5/32
+  - [501,476,  376,939,  749,711]  # P6/64
+  - [1126,489,  801,1222,  1618,1227]  # P7/128

models/hub/yolov3-spp.yaml

@@ -0,0 +1,51 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+
+# Parameters
+nc: 80  # number of classes
+depth_multiple: 1.0  # model depth multiple
+width_multiple: 1.0  # layer channel multiple
+anchors:
+  - [10,13, 16,30, 33,23]  # P3/8
+  - [30,61, 62,45, 59,119]  # P4/16
+  - [116,90, 156,198, 373,326]  # P5/32
+
+# darknet53 backbone
+backbone:
+  # [from, number, module, args]
+  [[-1, 1, Conv, [32, 3, 1]],  # 0
+   [-1, 1, Conv, [64, 3, 2]],  # 1-P1/2
+   [-1, 1, Bottleneck, [64]],
+   [-1, 1, Conv, [128, 3, 2]],  # 3-P2/4
+   [-1, 2, Bottleneck, [128]],
+   [-1, 1, Conv, [256, 3, 2]],  # 5-P3/8
+   [-1, 8, Bottleneck, [256]],
+   [-1, 1, Conv, [512, 3, 2]],  # 7-P4/16
+   [-1, 8, Bottleneck, [512]],
+   [-1, 1, Conv, [1024, 3, 2]],  # 9-P5/32
+   [-1, 4, Bottleneck, [1024]],  # 10
+  ]
+
+# YOLOv3-SPP head
+head:
+  [[-1, 1, Bottleneck, [1024, False]],
+   [-1, 1, SPP, [512, [5, 9, 13]]],
+   [-1, 1, Conv, [1024, 3, 1]],
+   [-1, 1, Conv, [512, 1, 1]],
+   [-1, 1, Conv, [1024, 3, 1]],  # 15 (P5/32-large)
+
+   [-2, 1, Conv, [256, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 8], 1, Concat, [1]],  # cat backbone P4
+   [-1, 1, Bottleneck, [512, False]],
+   [-1, 1, Bottleneck, [512, False]],
+   [-1, 1, Conv, [256, 1, 1]],
+   [-1, 1, Conv, [512, 3, 1]],  # 22 (P4/16-medium)
+
+   [-2, 1, Conv, [128, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 6], 1, Concat, [1]],  # cat backbone P3
+   [-1, 1, Bottleneck, [256, False]],
+   [-1, 2, Bottleneck, [256, False]],  # 27 (P3/8-small)
+
+   [[27, 22, 15], 1, Detect, [nc, anchors]],   # Detect(P3, P4, P5)
+  ]

models/hub/yolov3-tiny.yaml

@@ -0,0 +1,41 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+
+# Parameters
+nc: 80  # number of classes
+depth_multiple: 1.0  # model depth multiple
+width_multiple: 1.0  # layer channel multiple
+anchors:
+  - [10,14, 23,27, 37,58]  # P4/16
+  - [81,82, 135,169, 344,319]  # P5/32
+
+# YOLOv3-tiny backbone
+backbone:
+  # [from, number, module, args]
+  [[-1, 1, Conv, [16, 3, 1]],  # 0
+   [-1, 1, nn.MaxPool2d, [2, 2, 0]],  # 1-P1/2
+   [-1, 1, Conv, [32, 3, 1]],
+   [-1, 1, nn.MaxPool2d, [2, 2, 0]],  # 3-P2/4
+   [-1, 1, Conv, [64, 3, 1]],
+   [-1, 1, nn.MaxPool2d, [2, 2, 0]],  # 5-P3/8
+   [-1, 1, Conv, [128, 3, 1]],
+   [-1, 1, nn.MaxPool2d, [2, 2, 0]],  # 7-P4/16
+   [-1, 1, Conv, [256, 3, 1]],
+   [-1, 1, nn.MaxPool2d, [2, 2, 0]],  # 9-P5/32
+   [-1, 1, Conv, [512, 3, 1]],
+   [-1, 1, nn.ZeroPad2d, [[0, 1, 0, 1]]],  # 11
+   [-1, 1, nn.MaxPool2d, [2, 1, 0]],  # 12
+  ]
+
+# YOLOv3-tiny head
+head:
+  [[-1, 1, Conv, [1024, 3, 1]],
+   [-1, 1, Conv, [256, 1, 1]],
+   [-1, 1, Conv, [512, 3, 1]],  # 15 (P5/32-large)
+
+   [-2, 1, Conv, [128, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 8], 1, Concat, [1]],  # cat backbone P4
+   [-1, 1, Conv, [256, 3, 1]],  # 19 (P4/16-medium)
+
+   [[19, 15], 1, Detect, [nc, anchors]],  # Detect(P4, P5)
+  ]

models/hub/yolov3.yaml

@@ -0,0 +1,51 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+
+# Parameters
+nc: 80  # number of classes
+depth_multiple: 1.0  # model depth multiple
+width_multiple: 1.0  # layer channel multiple
+anchors:
+  - [10,13, 16,30, 33,23]  # P3/8
+  - [30,61, 62,45, 59,119]  # P4/16
+  - [116,90, 156,198, 373,326]  # P5/32
+
+# darknet53 backbone
+backbone:
+  # [from, number, module, args]
+  [[-1, 1, Conv, [32, 3, 1]],  # 0
+   [-1, 1, Conv, [64, 3, 2]],  # 1-P1/2
+   [-1, 1, Bottleneck, [64]],
+   [-1, 1, Conv, [128, 3, 2]],  # 3-P2/4
+   [-1, 2, Bottleneck, [128]],
+   [-1, 1, Conv, [256, 3, 2]],  # 5-P3/8
+   [-1, 8, Bottleneck, [256]],
+   [-1, 1, Conv, [512, 3, 2]],  # 7-P4/16
+   [-1, 8, Bottleneck, [512]],
+   [-1, 1, Conv, [1024, 3, 2]],  # 9-P5/32
+   [-1, 4, Bottleneck, [1024]],  # 10
+  ]
+
+# YOLOv3 head
+head:
+  [[-1, 1, Bottleneck, [1024, False]],
+   [-1, 1, Conv, [512, 1, 1]],
+   [-1, 1, Conv, [1024, 3, 1]],
+   [-1, 1, Conv, [512, 1, 1]],
+   [-1, 1, Conv, [1024, 3, 1]],  # 15 (P5/32-large)
+
+   [-2, 1, Conv, [256, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 8], 1, Concat, [1]],  # cat backbone P4
+   [-1, 1, Bottleneck, [512, False]],
+   [-1, 1, Bottleneck, [512, False]],
+   [-1, 1, Conv, [256, 1, 1]],
+   [-1, 1, Conv, [512, 3, 1]],  # 22 (P4/16-medium)
+
+   [-2, 1, Conv, [128, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 6], 1, Concat, [1]],  # cat backbone P3
+   [-1, 1, Bottleneck, [256, False]],
+   [-1, 2, Bottleneck, [256, False]],  # 27 (P3/8-small)
+
+   [[27, 22, 15], 1, Detect, [nc, anchors]],   # Detect(P3, P4, P5)
+  ]

models/hub/yolov5-bifpn.yaml

@@ -0,0 +1,48 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+
+# Parameters
+nc: 80  # number of classes
+depth_multiple: 1.0  # model depth multiple
+width_multiple: 1.0  # layer channel multiple
+anchors:
+  - [10,13, 16,30, 33,23]  # P3/8
+  - [30,61, 62,45, 59,119]  # P4/16
+  - [116,90, 156,198, 373,326]  # P5/32
+
+# YOLOv5 v6.0 backbone
+backbone:
+  # [from, number, module, args]
+  [[-1, 1, Conv, [64, 6, 2, 2]],  # 0-P1/2
+   [-1, 1, Conv, [128, 3, 2]],  # 1-P2/4
+   [-1, 3, C3, [128]],
+   [-1, 1, Conv, [256, 3, 2]],  # 3-P3/8
+   [-1, 6, C3, [256]],
+   [-1, 1, Conv, [512, 3, 2]],  # 5-P4/16
+   [-1, 9, C3, [512]],
+   [-1, 1, Conv, [1024, 3, 2]],  # 7-P5/32
+   [-1, 3, C3, [1024]],
+   [-1, 1, SPPF, [1024, 5]],  # 9
+  ]
+
+# YOLOv5 v6.0 BiFPN head
+head:
+  [[-1, 1, Conv, [512, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 6], 1, Concat, [1]],  # cat backbone P4
+   [-1, 3, C3, [512, False]],  # 13
+
+   [-1, 1, Conv, [256, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 4], 1, Concat, [1]],  # cat backbone P3
+   [-1, 3, C3, [256, False]],  # 17 (P3/8-small)
+
+   [-1, 1, Conv, [256, 3, 2]],
+   [[-1, 14, 6], 1, Concat, [1]],  # cat P4 <--- BiFPN change
+   [-1, 3, C3, [512, False]],  # 20 (P4/16-medium)
+
+   [-1, 1, Conv, [512, 3, 2]],
+   [[-1, 10], 1, Concat, [1]],  # cat head P5
+   [-1, 3, C3, [1024, False]],  # 23 (P5/32-large)
+
+   [[17, 20, 23], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5)
+  ]

models/hub/yolov5-fpn.yaml

@@ -0,0 +1,42 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+
+# Parameters
+nc: 80  # number of classes
+depth_multiple: 1.0  # model depth multiple
+width_multiple: 1.0  # layer channel multiple
+anchors:
+  - [10,13, 16,30, 33,23]  # P3/8
+  - [30,61, 62,45, 59,119]  # P4/16
+  - [116,90, 156,198, 373,326]  # P5/32
+
+# YOLOv5 v6.0 backbone
+backbone:
+  # [from, number, module, args]
+  [[-1, 1, Conv, [64, 6, 2, 2]],  # 0-P1/2
+   [-1, 1, Conv, [128, 3, 2]],  # 1-P2/4
+   [-1, 3, C3, [128]],
+   [-1, 1, Conv, [256, 3, 2]],  # 3-P3/8
+   [-1, 6, C3, [256]],
+   [-1, 1, Conv, [512, 3, 2]],  # 5-P4/16
+   [-1, 9, C3, [512]],
+   [-1, 1, Conv, [1024, 3, 2]],  # 7-P5/32
+   [-1, 3, C3, [1024]],
+   [-1, 1, SPPF, [1024, 5]],  # 9
+  ]
+
+# YOLOv5 v6.0 FPN head
+head:
+  [[-1, 3, C3, [1024, False]],  # 10 (P5/32-large)
+
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 6], 1, Concat, [1]],  # cat backbone P4
+   [-1, 1, Conv, [512, 1, 1]],
+   [-1, 3, C3, [512, False]],  # 14 (P4/16-medium)
+
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 4], 1, Concat, [1]],  # cat backbone P3
+   [-1, 1, Conv, [256, 1, 1]],
+   [-1, 3, C3, [256, False]],  # 18 (P3/8-small)
+
+   [[18, 14, 10], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5)
+  ]

models/hub/yolov5-p2.yaml

@@ -0,0 +1,54 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+
+# Parameters
+nc: 80  # number of classes
+depth_multiple: 1.0  # model depth multiple
+width_multiple: 1.0  # layer channel multiple
+anchors: 3  # AutoAnchor evolves 3 anchors per P output layer
+
+# YOLOv5 v6.0 backbone
+backbone:
+  # [from, number, module, args]
+  [[-1, 1, Conv, [64, 6, 2, 2]],  # 0-P1/2
+   [-1, 1, Conv, [128, 3, 2]],  # 1-P2/4
+   [-1, 3, C3, [128]],
+   [-1, 1, Conv, [256, 3, 2]],  # 3-P3/8
+   [-1, 6, C3, [256]],
+   [-1, 1, Conv, [512, 3, 2]],  # 5-P4/16
+   [-1, 9, C3, [512]],
+   [-1, 1, Conv, [1024, 3, 2]],  # 7-P5/32
+   [-1, 3, C3, [1024]],
+   [-1, 1, SPPF, [1024, 5]],  # 9
+  ]
+
+# YOLOv5 v6.0 head with (P2, P3, P4, P5) outputs
+head:
+  [[-1, 1, Conv, [512, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 6], 1, Concat, [1]],  # cat backbone P4
+   [-1, 3, C3, [512, False]],  # 13
+
+   [-1, 1, Conv, [256, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 4], 1, Concat, [1]],  # cat backbone P3
+   [-1, 3, C3, [256, False]],  # 17 (P3/8-small)
+
+   [-1, 1, Conv, [128, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 2], 1, Concat, [1]],  # cat backbone P2
+   [-1, 1, C3, [128, False]],  # 21 (P2/4-xsmall)
+
+   [-1, 1, Conv, [128, 3, 2]],
+   [[-1, 18], 1, Concat, [1]],  # cat head P3
+   [-1, 3, C3, [256, False]],  # 24 (P3/8-small)
+
+   [-1, 1, Conv, [256, 3, 2]],
+   [[-1, 14], 1, Concat, [1]],  # cat head P4
+   [-1, 3, C3, [512, False]],  # 27 (P4/16-medium)
+
+   [-1, 1, Conv, [512, 3, 2]],
+   [[-1, 10], 1, Concat, [1]],  # cat head P5
+   [-1, 3, C3, [1024, False]],  # 30 (P5/32-large)
+
+   [[21, 24, 27, 30], 1, Detect, [nc, anchors]],  # Detect(P2, P3, P4, P5)
+  ]

models/hub/yolov5-p34.yaml

@@ -0,0 +1,41 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+
+# Parameters
+nc: 80  # number of classes
+depth_multiple: 0.33  # model depth multiple
+width_multiple: 0.50  # layer channel multiple
+anchors: 3  # AutoAnchor evolves 3 anchors per P output layer
+
+# YOLOv5 v6.0 backbone
+backbone:
+  # [from, number, module, args]
+  [ [ -1, 1, Conv, [ 64, 6, 2, 2 ] ],  # 0-P1/2
+    [ -1, 1, Conv, [ 128, 3, 2 ] ],  # 1-P2/4
+    [ -1, 3, C3, [ 128 ] ],
+    [ -1, 1, Conv, [ 256, 3, 2 ] ],  # 3-P3/8
+    [ -1, 6, C3, [ 256 ] ],
+    [ -1, 1, Conv, [ 512, 3, 2 ] ],  # 5-P4/16
+    [ -1, 9, C3, [ 512 ] ],
+    [ -1, 1, Conv, [ 1024, 3, 2 ] ],  # 7-P5/32
+    [ -1, 3, C3, [ 1024 ] ],
+    [ -1, 1, SPPF, [ 1024, 5 ] ],  # 9
+  ]
+
+# YOLOv5 v6.0 head with (P3, P4) outputs
+head:
+  [ [ -1, 1, Conv, [ 512, 1, 1 ] ],
+    [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ],
+    [ [ -1, 6 ], 1, Concat, [ 1 ] ],  # cat backbone P4
+    [ -1, 3, C3, [ 512, False ] ],  # 13
+
+    [ -1, 1, Conv, [ 256, 1, 1 ] ],
+    [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ],
+    [ [ -1, 4 ], 1, Concat, [ 1 ] ],  # cat backbone P3
+    [ -1, 3, C3, [ 256, False ] ],  # 17 (P3/8-small)
+
+    [ -1, 1, Conv, [ 256, 3, 2 ] ],
+    [ [ -1, 14 ], 1, Concat, [ 1 ] ],  # cat head P4
+    [ -1, 3, C3, [ 512, False ] ],  # 20 (P4/16-medium)
+
+    [ [ 17, 20 ], 1, Detect, [ nc, anchors ] ],  # Detect(P3, P4)
+  ]

models/hub/yolov5-p6.yaml

@@ -0,0 +1,56 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+
+# Parameters
+nc: 80  # number of classes
+depth_multiple: 1.0  # model depth multiple
+width_multiple: 1.0  # layer channel multiple
+anchors: 3  # AutoAnchor evolves 3 anchors per P output layer
+
+# YOLOv5 v6.0 backbone
+backbone:
+  # [from, number, module, args]
+  [[-1, 1, Conv, [64, 6, 2, 2]],  # 0-P1/2
+   [-1, 1, Conv, [128, 3, 2]],  # 1-P2/4
+   [-1, 3, C3, [128]],
+   [-1, 1, Conv, [256, 3, 2]],  # 3-P3/8
+   [-1, 6, C3, [256]],
+   [-1, 1, Conv, [512, 3, 2]],  # 5-P4/16
+   [-1, 9, C3, [512]],
+   [-1, 1, Conv, [768, 3, 2]],  # 7-P5/32
+   [-1, 3, C3, [768]],
+   [-1, 1, Conv, [1024, 3, 2]],  # 9-P6/64
+   [-1, 3, C3, [1024]],
+   [-1, 1, SPPF, [1024, 5]],  # 11
+  ]
+
+# YOLOv5 v6.0 head with (P3, P4, P5, P6) outputs
+head:
+  [[-1, 1, Conv, [768, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 8], 1, Concat, [1]],  # cat backbone P5
+   [-1, 3, C3, [768, False]],  # 15
+
+   [-1, 1, Conv, [512, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 6], 1, Concat, [1]],  # cat backbone P4
+   [-1, 3, C3, [512, False]],  # 19
+
+   [-1, 1, Conv, [256, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 4], 1, Concat, [1]],  # cat backbone P3
+   [-1, 3, C3, [256, False]],  # 23 (P3/8-small)
+
+   [-1, 1, Conv, [256, 3, 2]],
+   [[-1, 20], 1, Concat, [1]],  # cat head P4
+   [-1, 3, C3, [512, False]],  # 26 (P4/16-medium)
+
+   [-1, 1, Conv, [512, 3, 2]],
+   [[-1, 16], 1, Concat, [1]],  # cat head P5
+   [-1, 3, C3, [768, False]],  # 29 (P5/32-large)
+
+   [-1, 1, Conv, [768, 3, 2]],
+   [[-1, 12], 1, Concat, [1]],  # cat head P6
+   [-1, 3, C3, [1024, False]],  # 32 (P6/64-xlarge)
+
+   [[23, 26, 29, 32], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5, P6)
+  ]

models/hub/yolov5-p7.yaml

@@ -0,0 +1,67 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+
+# Parameters
+nc: 80  # number of classes
+depth_multiple: 1.0  # model depth multiple
+width_multiple: 1.0  # layer channel multiple
+anchors: 3  # AutoAnchor evolves 3 anchors per P output layer
+
+# YOLOv5 v6.0 backbone
+backbone:
+  # [from, number, module, args]
+  [[-1, 1, Conv, [64, 6, 2, 2]],  # 0-P1/2
+   [-1, 1, Conv, [128, 3, 2]],  # 1-P2/4
+   [-1, 3, C3, [128]],
+   [-1, 1, Conv, [256, 3, 2]],  # 3-P3/8
+   [-1, 6, C3, [256]],
+   [-1, 1, Conv, [512, 3, 2]],  # 5-P4/16
+   [-1, 9, C3, [512]],
+   [-1, 1, Conv, [768, 3, 2]],  # 7-P5/32
+   [-1, 3, C3, [768]],
+   [-1, 1, Conv, [1024, 3, 2]],  # 9-P6/64
+   [-1, 3, C3, [1024]],
+   [-1, 1, Conv, [1280, 3, 2]],  # 11-P7/128
+   [-1, 3, C3, [1280]],
+   [-1, 1, SPPF, [1280, 5]],  # 13
+  ]
+
+# YOLOv5 v6.0 head with (P3, P4, P5, P6, P7) outputs
+head:
+  [[-1, 1, Conv, [1024, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 10], 1, Concat, [1]],  # cat backbone P6
+   [-1, 3, C3, [1024, False]],  # 17
+
+   [-1, 1, Conv, [768, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 8], 1, Concat, [1]],  # cat backbone P5
+   [-1, 3, C3, [768, False]],  # 21
+
+   [-1, 1, Conv, [512, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 6], 1, Concat, [1]],  # cat backbone P4
+   [-1, 3, C3, [512, False]],  # 25
+
+   [-1, 1, Conv, [256, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 4], 1, Concat, [1]],  # cat backbone P3
+   [-1, 3, C3, [256, False]],  # 29 (P3/8-small)
+
+   [-1, 1, Conv, [256, 3, 2]],
+   [[-1, 26], 1, Concat, [1]],  # cat head P4
+   [-1, 3, C3, [512, False]],  # 32 (P4/16-medium)
+
+   [-1, 1, Conv, [512, 3, 2]],
+   [[-1, 22], 1, Concat, [1]],  # cat head P5
+   [-1, 3, C3, [768, False]],  # 35 (P5/32-large)
+
+   [-1, 1, Conv, [768, 3, 2]],
+   [[-1, 18], 1, Concat, [1]],  # cat head P6
+   [-1, 3, C3, [1024, False]],  # 38 (P6/64-xlarge)
+
+   [-1, 1, Conv, [1024, 3, 2]],
+   [[-1, 14], 1, Concat, [1]],  # cat head P7
+   [-1, 3, C3, [1280, False]],  # 41 (P7/128-xxlarge)
+
+   [[29, 32, 35, 38, 41], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5, P6, P7)
+  ]

models/hub/yolov5-panet.yaml

@@ -0,0 +1,48 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+
+# Parameters
+nc: 80  # number of classes
+depth_multiple: 1.0  # model depth multiple
+width_multiple: 1.0  # layer channel multiple
+anchors:
+  - [10,13, 16,30, 33,23]  # P3/8
+  - [30,61, 62,45, 59,119]  # P4/16
+  - [116,90, 156,198, 373,326]  # P5/32
+
+# YOLOv5 v6.0 backbone
+backbone:
+  # [from, number, module, args]
+  [[-1, 1, Conv, [64, 6, 2, 2]],  # 0-P1/2
+   [-1, 1, Conv, [128, 3, 2]],  # 1-P2/4
+   [-1, 3, C3, [128]],
+   [-1, 1, Conv, [256, 3, 2]],  # 3-P3/8
+   [-1, 6, C3, [256]],
+   [-1, 1, Conv, [512, 3, 2]],  # 5-P4/16
+   [-1, 9, C3, [512]],
+   [-1, 1, Conv, [1024, 3, 2]],  # 7-P5/32
+   [-1, 3, C3, [1024]],
+   [-1, 1, SPPF, [1024, 5]],  # 9
+  ]
+
+# YOLOv5 v6.0 PANet head
+head:
+  [[-1, 1, Conv, [512, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 6], 1, Concat, [1]],  # cat backbone P4
+   [-1, 3, C3, [512, False]],  # 13
+
+   [-1, 1, Conv, [256, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 4], 1, Concat, [1]],  # cat backbone P3
+   [-1, 3, C3, [256, False]],  # 17 (P3/8-small)
+
+   [-1, 1, Conv, [256, 3, 2]],
+   [[-1, 14], 1, Concat, [1]],  # cat head P4
+   [-1, 3, C3, [512, False]],  # 20 (P4/16-medium)
+
+   [-1, 1, Conv, [512, 3, 2]],
+   [[-1, 10], 1, Concat, [1]],  # cat head P5
+   [-1, 3, C3, [1024, False]],  # 23 (P5/32-large)
+
+   [[17, 20, 23], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5)
+  ]

models/hub/yolov5l6.yaml

@@ -0,0 +1,60 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+
+# Parameters
+nc: 80  # number of classes
+depth_multiple: 1.0  # model depth multiple
+width_multiple: 1.0  # layer channel multiple
+anchors:
+  - [19,27,  44,40,  38,94]  # P3/8
+  - [96,68,  86,152,  180,137]  # P4/16
+  - [140,301,  303,264,  238,542]  # P5/32
+  - [436,615,  739,380,  925,792]  # P6/64
+
+# YOLOv5 v6.0 backbone
+backbone:
+  # [from, number, module, args]
+  [[-1, 1, Conv, [64, 6, 2, 2]],  # 0-P1/2
+   [-1, 1, Conv, [128, 3, 2]],  # 1-P2/4
+   [-1, 3, C3, [128]],
+   [-1, 1, Conv, [256, 3, 2]],  # 3-P3/8
+   [-1, 6, C3, [256]],
+   [-1, 1, Conv, [512, 3, 2]],  # 5-P4/16
+   [-1, 9, C3, [512]],
+   [-1, 1, Conv, [768, 3, 2]],  # 7-P5/32
+   [-1, 3, C3, [768]],
+   [-1, 1, Conv, [1024, 3, 2]],  # 9-P6/64
+   [-1, 3, C3, [1024]],
+   [-1, 1, SPPF, [1024, 5]],  # 11
+  ]
+
+# YOLOv5 v6.0 head
+head:
+  [[-1, 1, Conv, [768, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 8], 1, Concat, [1]],  # cat backbone P5
+   [-1, 3, C3, [768, False]],  # 15
+
+   [-1, 1, Conv, [512, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 6], 1, Concat, [1]],  # cat backbone P4
+   [-1, 3, C3, [512, False]],  # 19
+
+   [-1, 1, Conv, [256, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 4], 1, Concat, [1]],  # cat backbone P3
+   [-1, 3, C3, [256, False]],  # 23 (P3/8-small)
+
+   [-1, 1, Conv, [256, 3, 2]],
+   [[-1, 20], 1, Concat, [1]],  # cat head P4
+   [-1, 3, C3, [512, False]],  # 26 (P4/16-medium)
+
+   [-1, 1, Conv, [512, 3, 2]],
+   [[-1, 16], 1, Concat, [1]],  # cat head P5
+   [-1, 3, C3, [768, False]],  # 29 (P5/32-large)
+
+   [-1, 1, Conv, [768, 3, 2]],
+   [[-1, 12], 1, Concat, [1]],  # cat head P6
+   [-1, 3, C3, [1024, False]],  # 32 (P6/64-xlarge)
+
+   [[23, 26, 29, 32], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5, P6)
+  ]

models/hub/yolov5m6.yaml

@@ -0,0 +1,60 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+
+# Parameters
+nc: 80  # number of classes
+depth_multiple: 0.67  # model depth multiple
+width_multiple: 0.75  # layer channel multiple
+anchors:
+  - [19,27,  44,40,  38,94]  # P3/8
+  - [96,68,  86,152,  180,137]  # P4/16
+  - [140,301,  303,264,  238,542]  # P5/32
+  - [436,615,  739,380,  925,792]  # P6/64
+
+# YOLOv5 v6.0 backbone
+backbone:
+  # [from, number, module, args]
+  [[-1, 1, Conv, [64, 6, 2, 2]],  # 0-P1/2
+   [-1, 1, Conv, [128, 3, 2]],  # 1-P2/4
+   [-1, 3, C3, [128]],
+   [-1, 1, Conv, [256, 3, 2]],  # 3-P3/8
+   [-1, 6, C3, [256]],
+   [-1, 1, Conv, [512, 3, 2]],  # 5-P4/16
+   [-1, 9, C3, [512]],
+   [-1, 1, Conv, [768, 3, 2]],  # 7-P5/32
+   [-1, 3, C3, [768]],
+   [-1, 1, Conv, [1024, 3, 2]],  # 9-P6/64
+   [-1, 3, C3, [1024]],
+   [-1, 1, SPPF, [1024, 5]],  # 11
+  ]
+
+# YOLOv5 v6.0 head
+head:
+  [[-1, 1, Conv, [768, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 8], 1, Concat, [1]],  # cat backbone P5
+   [-1, 3, C3, [768, False]],  # 15
+
+   [-1, 1, Conv, [512, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 6], 1, Concat, [1]],  # cat backbone P4
+   [-1, 3, C3, [512, False]],  # 19
+
+   [-1, 1, Conv, [256, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 4], 1, Concat, [1]],  # cat backbone P3
+   [-1, 3, C3, [256, False]],  # 23 (P3/8-small)
+
+   [-1, 1, Conv, [256, 3, 2]],
+   [[-1, 20], 1, Concat, [1]],  # cat head P4
+   [-1, 3, C3, [512, False]],  # 26 (P4/16-medium)
+
+   [-1, 1, Conv, [512, 3, 2]],
+   [[-1, 16], 1, Concat, [1]],  # cat head P5
+   [-1, 3, C3, [768, False]],  # 29 (P5/32-large)
+
+   [-1, 1, Conv, [768, 3, 2]],
+   [[-1, 12], 1, Concat, [1]],  # cat head P6
+   [-1, 3, C3, [1024, False]],  # 32 (P6/64-xlarge)
+
+   [[23, 26, 29, 32], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5, P6)
+  ]

models/hub/yolov5n6.yaml

@@ -0,0 +1,60 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+
+# Parameters
+nc: 80  # number of classes
+depth_multiple: 0.33  # model depth multiple
+width_multiple: 0.25  # layer channel multiple
+anchors:
+  - [19,27,  44,40,  38,94]  # P3/8
+  - [96,68,  86,152,  180,137]  # P4/16
+  - [140,301,  303,264,  238,542]  # P5/32
+  - [436,615,  739,380,  925,792]  # P6/64
+
+# YOLOv5 v6.0 backbone
+backbone:
+  # [from, number, module, args]
+  [[-1, 1, Conv, [64, 6, 2, 2]],  # 0-P1/2
+   [-1, 1, Conv, [128, 3, 2]],  # 1-P2/4
+   [-1, 3, C3, [128]],
+   [-1, 1, Conv, [256, 3, 2]],  # 3-P3/8
+   [-1, 6, C3, [256]],
+   [-1, 1, Conv, [512, 3, 2]],  # 5-P4/16
+   [-1, 9, C3, [512]],
+   [-1, 1, Conv, [768, 3, 2]],  # 7-P5/32
+   [-1, 3, C3, [768]],
+   [-1, 1, Conv, [1024, 3, 2]],  # 9-P6/64
+   [-1, 3, C3, [1024]],
+   [-1, 1, SPPF, [1024, 5]],  # 11
+  ]
+
+# YOLOv5 v6.0 head
+head:
+  [[-1, 1, Conv, [768, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 8], 1, Concat, [1]],  # cat backbone P5
+   [-1, 3, C3, [768, False]],  # 15
+
+   [-1, 1, Conv, [512, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 6], 1, Concat, [1]],  # cat backbone P4
+   [-1, 3, C3, [512, False]],  # 19
+
+   [-1, 1, Conv, [256, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 4], 1, Concat, [1]],  # cat backbone P3
+   [-1, 3, C3, [256, False]],  # 23 (P3/8-small)
+
+   [-1, 1, Conv, [256, 3, 2]],
+   [[-1, 20], 1, Concat, [1]],  # cat head P4
+   [-1, 3, C3, [512, False]],  # 26 (P4/16-medium)
+
+   [-1, 1, Conv, [512, 3, 2]],
+   [[-1, 16], 1, Concat, [1]],  # cat head P5
+   [-1, 3, C3, [768, False]],  # 29 (P5/32-large)
+
+   [-1, 1, Conv, [768, 3, 2]],
+   [[-1, 12], 1, Concat, [1]],  # cat head P6
+   [-1, 3, C3, [1024, False]],  # 32 (P6/64-xlarge)
+
+   [[23, 26, 29, 32], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5, P6)
+  ]

models/hub/yolov5s-ghost.yaml

@@ -0,0 +1,48 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+
+# Parameters
+nc: 80  # number of classes
+depth_multiple: 0.33  # model depth multiple
+width_multiple: 0.50  # layer channel multiple
+anchors:
+  - [10,13, 16,30, 33,23]  # P3/8
+  - [30,61, 62,45, 59,119]  # P4/16
+  - [116,90, 156,198, 373,326]  # P5/32
+
+# YOLOv5 v6.0 backbone
+backbone:
+  # [from, number, module, args]
+  [[-1, 1, Conv, [64, 6, 2, 2]],  # 0-P1/2
+   [-1, 1, GhostConv, [128, 3, 2]],  # 1-P2/4
+   [-1, 3, C3Ghost, [128]],
+   [-1, 1, GhostConv, [256, 3, 2]],  # 3-P3/8
+   [-1, 6, C3Ghost, [256]],
+   [-1, 1, GhostConv, [512, 3, 2]],  # 5-P4/16
+   [-1, 9, C3Ghost, [512]],
+   [-1, 1, GhostConv, [1024, 3, 2]],  # 7-P5/32
+   [-1, 3, C3Ghost, [1024]],
+   [-1, 1, SPPF, [1024, 5]],  # 9
+  ]
+
+# YOLOv5 v6.0 head
+head:
+  [[-1, 1, GhostConv, [512, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 6], 1, Concat, [1]],  # cat backbone P4
+   [-1, 3, C3Ghost, [512, False]],  # 13
+
+   [-1, 1, GhostConv, [256, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 4], 1, Concat, [1]],  # cat backbone P3
+   [-1, 3, C3Ghost, [256, False]],  # 17 (P3/8-small)
+
+   [-1, 1, GhostConv, [256, 3, 2]],
+   [[-1, 14], 1, Concat, [1]],  # cat head P4
+   [-1, 3, C3Ghost, [512, False]],  # 20 (P4/16-medium)
+
+   [-1, 1, GhostConv, [512, 3, 2]],
+   [[-1, 10], 1, Concat, [1]],  # cat head P5
+   [-1, 3, C3Ghost, [1024, False]],  # 23 (P5/32-large)
+
+   [[17, 20, 23], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5)
+  ]

models/hub/yolov5s-transformer.yaml

@@ -0,0 +1,48 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+
+# Parameters
+nc: 80  # number of classes
+depth_multiple: 0.33  # model depth multiple
+width_multiple: 0.50  # layer channel multiple
+anchors:
+  - [10,13, 16,30, 33,23]  # P3/8
+  - [30,61, 62,45, 59,119]  # P4/16
+  - [116,90, 156,198, 373,326]  # P5/32
+
+# YOLOv5 v6.0 backbone
+backbone:
+  # [from, number, module, args]
+  [[-1, 1, Conv, [64, 6, 2, 2]],  # 0-P1/2
+   [-1, 1, Conv, [128, 3, 2]],  # 1-P2/4
+   [-1, 3, C3, [128]],
+   [-1, 1, Conv, [256, 3, 2]],  # 3-P3/8
+   [-1, 6, C3, [256]],
+   [-1, 1, Conv, [512, 3, 2]],  # 5-P4/16
+   [-1, 9, C3, [512]],
+   [-1, 1, Conv, [1024, 3, 2]],  # 7-P5/32
+   [-1, 3, C3TR, [1024]],  # 9 <--- C3TR() Transformer module
+   [-1, 1, SPPF, [1024, 5]],  # 9
+  ]
+
+# YOLOv5 v6.0 head
+head:
+  [[-1, 1, Conv, [512, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 6], 1, Concat, [1]],  # cat backbone P4
+   [-1, 3, C3, [512, False]],  # 13
+
+   [-1, 1, Conv, [256, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 4], 1, Concat, [1]],  # cat backbone P3
+   [-1, 3, C3, [256, False]],  # 17 (P3/8-small)
+
+   [-1, 1, Conv, [256, 3, 2]],
+   [[-1, 14], 1, Concat, [1]],  # cat head P4
+   [-1, 3, C3, [512, False]],  # 20 (P4/16-medium)
+
+   [-1, 1, Conv, [512, 3, 2]],
+   [[-1, 10], 1, Concat, [1]],  # cat head P5
+   [-1, 3, C3, [1024, False]],  # 23 (P5/32-large)
+
+   [[17, 20, 23], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5)
+  ]

models/hub/yolov5s6.yaml

@@ -0,0 +1,60 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+
+# Parameters
+nc: 80  # number of classes
+depth_multiple: 0.33  # model depth multiple
+width_multiple: 0.50  # layer channel multiple
+anchors:
+  - [19,27,  44,40,  38,94]  # P3/8
+  - [96,68,  86,152,  180,137]  # P4/16
+  - [140,301,  303,264,  238,542]  # P5/32
+  - [436,615,  739,380,  925,792]  # P6/64
+
+# YOLOv5 v6.0 backbone
+backbone:
+  # [from, number, module, args]
+  [[-1, 1, Conv, [64, 6, 2, 2]],  # 0-P1/2
+   [-1, 1, Conv, [128, 3, 2]],  # 1-P2/4
+   [-1, 3, C3, [128]],
+   [-1, 1, Conv, [256, 3, 2]],  # 3-P3/8
+   [-1, 6, C3, [256]],
+   [-1, 1, Conv, [512, 3, 2]],  # 5-P4/16
+   [-1, 9, C3, [512]],
+   [-1, 1, Conv, [768, 3, 2]],  # 7-P5/32
+   [-1, 3, C3, [768]],
+   [-1, 1, Conv, [1024, 3, 2]],  # 9-P6/64
+   [-1, 3, C3, [1024]],
+   [-1, 1, SPPF, [1024, 5]],  # 11
+  ]
+
+# YOLOv5 v6.0 head
+head:
+  [[-1, 1, Conv, [768, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 8], 1, Concat, [1]],  # cat backbone P5
+   [-1, 3, C3, [768, False]],  # 15
+
+   [-1, 1, Conv, [512, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 6], 1, Concat, [1]],  # cat backbone P4
+   [-1, 3, C3, [512, False]],  # 19
+
+   [-1, 1, Conv, [256, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 4], 1, Concat, [1]],  # cat backbone P3
+   [-1, 3, C3, [256, False]],  # 23 (P3/8-small)
+
+   [-1, 1, Conv, [256, 3, 2]],
+   [[-1, 20], 1, Concat, [1]],  # cat head P4
+   [-1, 3, C3, [512, False]],  # 26 (P4/16-medium)
+
+   [-1, 1, Conv, [512, 3, 2]],
+   [[-1, 16], 1, Concat, [1]],  # cat head P5
+   [-1, 3, C3, [768, False]],  # 29 (P5/32-large)
+
+   [-1, 1, Conv, [768, 3, 2]],
+   [[-1, 12], 1, Concat, [1]],  # cat head P6
+   [-1, 3, C3, [1024, False]],  # 32 (P6/64-xlarge)
+
+   [[23, 26, 29, 32], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5, P6)
+  ]

models/hub/yolov5x6.yaml

@@ -0,0 +1,60 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+
+# Parameters
+nc: 80  # number of classes
+depth_multiple: 1.33  # model depth multiple
+width_multiple: 1.25  # layer channel multiple
+anchors:
+  - [19,27,  44,40,  38,94]  # P3/8
+  - [96,68,  86,152,  180,137]  # P4/16
+  - [140,301,  303,264,  238,542]  # P5/32
+  - [436,615,  739,380,  925,792]  # P6/64
+
+# YOLOv5 v6.0 backbone
+backbone:
+  # [from, number, module, args]
+  [[-1, 1, Conv, [64, 6, 2, 2]],  # 0-P1/2
+   [-1, 1, Conv, [128, 3, 2]],  # 1-P2/4
+   [-1, 3, C3, [128]],
+   [-1, 1, Conv, [256, 3, 2]],  # 3-P3/8
+   [-1, 6, C3, [256]],
+   [-1, 1, Conv, [512, 3, 2]],  # 5-P4/16
+   [-1, 9, C3, [512]],
+   [-1, 1, Conv, [768, 3, 2]],  # 7-P5/32
+   [-1, 3, C3, [768]],
+   [-1, 1, Conv, [1024, 3, 2]],  # 9-P6/64
+   [-1, 3, C3, [1024]],
+   [-1, 1, SPPF, [1024, 5]],  # 11
+  ]
+
+# YOLOv5 v6.0 head
+head:
+  [[-1, 1, Conv, [768, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 8], 1, Concat, [1]],  # cat backbone P5
+   [-1, 3, C3, [768, False]],  # 15
+
+   [-1, 1, Conv, [512, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 6], 1, Concat, [1]],  # cat backbone P4
+   [-1, 3, C3, [512, False]],  # 19
+
+   [-1, 1, Conv, [256, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 4], 1, Concat, [1]],  # cat backbone P3
+   [-1, 3, C3, [256, False]],  # 23 (P3/8-small)
+
+   [-1, 1, Conv, [256, 3, 2]],
+   [[-1, 20], 1, Concat, [1]],  # cat head P4
+   [-1, 3, C3, [512, False]],  # 26 (P4/16-medium)
+
+   [-1, 1, Conv, [512, 3, 2]],
+   [[-1, 16], 1, Concat, [1]],  # cat head P5
+   [-1, 3, C3, [768, False]],  # 29 (P5/32-large)
+
+   [-1, 1, Conv, [768, 3, 2]],
+   [[-1, 12], 1, Concat, [1]],  # cat head P6
+   [-1, 3, C3, [1024, False]],  # 32 (P6/64-xlarge)
+
+   [[23, 26, 29, 32], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5, P6)
+  ]

models/pruned_common.py

@@ -0,0 +1,69 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+"""
+Common modules
+"""
+
+import json
+import math
+import platform
+import warnings
+from collections import OrderedDict, namedtuple
+from copy import copy
+from pathlib import Path
+
+import cv2
+import numpy as np
+import pandas as pd
+import requests
+import torch
+import torch.nn as nn
+from PIL import Image
+from torch.cuda import amp
+
+from utils.datasets import exif_transpose, letterbox
+from utils.general import (LOGGER, check_requirements, check_suffix, check_version, colorstr, increment_path,
+                           make_divisible, non_max_suppression, scale_coords, xywh2xyxy, xyxy2xywh)
+from utils.plots import Annotator, colors, save_one_box
+from utils.torch_utils import copy_attr, time_sync
+from models.common import Conv
+
+class BottleneckPruned(nn.Module):
+    # Pruned bottleneck
+    def __init__(self, cv1in, cv1out, cv2out, shortcut=True, g=1):  # ch_in, ch_out, shortcut, groups, expansion
+        super(BottleneckPruned, self).__init__()
+        self.cv1 = Conv(cv1in, cv1out, 1, 1)
+        self.cv2 = Conv(cv1out, cv2out, 3, 1, g=g)
+        self.add = shortcut and cv1in == cv2out
+
+    def forward(self, x):
+        return x + self.cv2(self.cv1(x)) if self.add else self.cv2(self.cv1(x))
+
+class C3Pruned(nn.Module):
+    # CSP Bottleneck with 3 convolutions
+    def __init__(self, cv1in, cv1out, cv2out, cv3out, bottle_args, n=1, shortcut=True, g=1):  # ch_in, ch_out, number, shortcut, groups, expansion
+        super(C3Pruned, self).__init__()
+        cv3in = bottle_args[-1][-1]
+        self.cv1 = Conv(cv1in, cv1out, 1, 1)
+        self.cv2 = Conv(cv1in, cv2out, 1, 1)
+        self.cv3 = Conv(cv3in+cv2out, cv3out, 1)
+        self.m = nn.Sequential(*[BottleneckPruned(*bottle_args[k], shortcut, g) for k in range(n)])
+
+    def forward(self, x):
+        return self.cv3(torch.cat((self.m(self.cv1(x)), self.cv2(x)), dim=1))
+
+
+class SPPFPruned(nn.Module):
+    # Spatial pyramid pooling layer used in YOLOv3-SPP
+    def __init__(self, cv1in, cv1out, cv2out, k=5):
+        super(SPPFPruned, self).__init__()
+        self.cv1 = Conv(cv1in, cv1out, 1, 1)
+        self.cv2 = Conv(cv1out * 4, cv2out, 1, 1)
+        self.m = nn.MaxPool2d(kernel_size=k, stride=1, padding=k // 2)
+
+    def forward(self, x):
+        x = self.cv1(x)
+        with warnings.catch_warnings():
+            warnings.simplefilter('ignore')  # suppress torch 1.9.0 max_pool2d() warning
+            y1 = self.m(x)
+            y2 = self.m(y1)
+            return self.cv2(torch.cat([x, y1, y2, self.m(y2)], 1))

models/tf.py

@@ -0,0 +1,464 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+"""
+TensorFlow, Keras and TFLite versions of YOLOv5
+Authored by https://github.com/zldrobit in PR https://github.com/ultralytics/yolov5/pull/1127
+
+Usage:
+    $ python models/tf.py --weights yolov5s.pt
+
+Export:
+    $ python path/to/export.py --weights yolov5s.pt --include saved_model pb tflite tfjs
+"""
+
+import argparse
+import sys
+from copy import deepcopy
+from pathlib import Path
+
+FILE = Path(__file__).resolve()
+ROOT = FILE.parents[1]  # YOLOv5 root directory
+if str(ROOT) not in sys.path:
+    sys.path.append(str(ROOT))  # add ROOT to PATH
+# ROOT = ROOT.relative_to(Path.cwd())  # relative
+
+import numpy as np
+import tensorflow as tf
+import torch
+import torch.nn as nn
+from tensorflow import keras
+
+from models.common import C3, SPP, SPPF, Bottleneck, BottleneckCSP, Concat, Conv, DWConv, Focus, autopad
+from models.experimental import CrossConv, MixConv2d, attempt_load
+from models.yolo import Detect
+from utils.activations import SiLU
+from utils.general import LOGGER, make_divisible, print_args
+
+
+class TFBN(keras.layers.Layer):
+    # TensorFlow BatchNormalization wrapper
+    def __init__(self, w=None):
+        super().__init__()
+        self.bn = keras.layers.BatchNormalization(
+            beta_initializer=keras.initializers.Constant(w.bias.numpy()),
+            gamma_initializer=keras.initializers.Constant(w.weight.numpy()),
+            moving_mean_initializer=keras.initializers.Constant(w.running_mean.numpy()),
+            moving_variance_initializer=keras.initializers.Constant(w.running_var.numpy()),
+            epsilon=w.eps)
+
+    def call(self, inputs):
+        return self.bn(inputs)
+
+
+class TFPad(keras.layers.Layer):
+    def __init__(self, pad):
+        super().__init__()
+        self.pad = tf.constant([[0, 0], [pad, pad], [pad, pad], [0, 0]])
+
+    def call(self, inputs):
+        return tf.pad(inputs, self.pad, mode='constant', constant_values=0)
+
+
+class TFConv(keras.layers.Layer):
+    # Standard convolution
+    def __init__(self, c1, c2, k=1, s=1, p=None, g=1, act=True, w=None):
+        # ch_in, ch_out, weights, kernel, stride, padding, groups
+        super().__init__()
+        assert g == 1, "TF v2.2 Conv2D does not support 'groups' argument"
+        assert isinstance(k, int), "Convolution with multiple kernels are not allowed."
+        # TensorFlow convolution padding is inconsistent with PyTorch (e.g. k=3 s=2 'SAME' padding)
+        # see https://stackoverflow.com/questions/52975843/comparing-conv2d-with-padding-between-tensorflow-and-pytorch
+
+        conv = keras.layers.Conv2D(
+            c2, k, s, 'SAME' if s == 1 else 'VALID', use_bias=False if hasattr(w, 'bn') else True,
+            kernel_initializer=keras.initializers.Constant(w.conv.weight.permute(2, 3, 1, 0).numpy()),
+            bias_initializer='zeros' if hasattr(w, 'bn') else keras.initializers.Constant(w.conv.bias.numpy()))
+        self.conv = conv if s == 1 else keras.Sequential([TFPad(autopad(k, p)), conv])
+        self.bn = TFBN(w.bn) if hasattr(w, 'bn') else tf.identity
+
+        # YOLOv5 activations
+        if isinstance(w.act, nn.LeakyReLU):
+            self.act = (lambda x: keras.activations.relu(x, alpha=0.1)) if act else tf.identity
+        elif isinstance(w.act, nn.Hardswish):
+            self.act = (lambda x: x * tf.nn.relu6(x + 3) * 0.166666667) if act else tf.identity
+        elif isinstance(w.act, (nn.SiLU, SiLU)):
+            self.act = (lambda x: keras.activations.swish(x)) if act else tf.identity
+        else:
+            raise Exception(f'no matching TensorFlow activation found for {w.act}')
+
+    def call(self, inputs):
+        return self.act(self.bn(self.conv(inputs)))
+
+
+class TFFocus(keras.layers.Layer):
+    # Focus wh information into c-space
+    def __init__(self, c1, c2, k=1, s=1, p=None, g=1, act=True, w=None):
+        # ch_in, ch_out, kernel, stride, padding, groups
+        super().__init__()
+        self.conv = TFConv(c1 * 4, c2, k, s, p, g, act, w.conv)
+
+    def call(self, inputs):  # x(b,w,h,c) -> y(b,w/2,h/2,4c)
+        # inputs = inputs / 255  # normalize 0-255 to 0-1
+        return self.conv(tf.concat([inputs[:, ::2, ::2, :],
+                                    inputs[:, 1::2, ::2, :],
+                                    inputs[:, ::2, 1::2, :],
+                                    inputs[:, 1::2, 1::2, :]], 3))
+
+
+class TFBottleneck(keras.layers.Layer):
+    # Standard bottleneck
+    def __init__(self, c1, c2, shortcut=True, g=1, e=0.5, w=None):  # ch_in, ch_out, shortcut, groups, expansion
+        super().__init__()
+        c_ = int(c2 * e)  # hidden channels
+        self.cv1 = TFConv(c1, c_, 1, 1, w=w.cv1)
+        self.cv2 = TFConv(c_, c2, 3, 1, g=g, w=w.cv2)
+        self.add = shortcut and c1 == c2
+
+    def call(self, inputs):
+        return inputs + self.cv2(self.cv1(inputs)) if self.add else self.cv2(self.cv1(inputs))
+
+
+class TFConv2d(keras.layers.Layer):
+    # Substitution for PyTorch nn.Conv2D
+    def __init__(self, c1, c2, k, s=1, g=1, bias=True, w=None):
+        super().__init__()
+        assert g == 1, "TF v2.2 Conv2D does not support 'groups' argument"
+        self.conv = keras.layers.Conv2D(
+            c2, k, s, 'VALID', use_bias=bias,
+            kernel_initializer=keras.initializers.Constant(w.weight.permute(2, 3, 1, 0).numpy()),
+            bias_initializer=keras.initializers.Constant(w.bias.numpy()) if bias else None, )
+
+    def call(self, inputs):
+        return self.conv(inputs)
+
+
+class TFBottleneckCSP(keras.layers.Layer):
+    # CSP Bottleneck https://github.com/WongKinYiu/CrossStagePartialNetworks
+    def __init__(self, c1, c2, n=1, shortcut=True, g=1, e=0.5, w=None):
+        # ch_in, ch_out, number, shortcut, groups, expansion
+        super().__init__()
+        c_ = int(c2 * e)  # hidden channels
+        self.cv1 = TFConv(c1, c_, 1, 1, w=w.cv1)
+        self.cv2 = TFConv2d(c1, c_, 1, 1, bias=False, w=w.cv2)
+        self.cv3 = TFConv2d(c_, c_, 1, 1, bias=False, w=w.cv3)
+        self.cv4 = TFConv(2 * c_, c2, 1, 1, w=w.cv4)
+        self.bn = TFBN(w.bn)
+        self.act = lambda x: keras.activations.relu(x, alpha=0.1)
+        self.m = keras.Sequential([TFBottleneck(c_, c_, shortcut, g, e=1.0, w=w.m[j]) for j in range(n)])
+
+    def call(self, inputs):
+        y1 = self.cv3(self.m(self.cv1(inputs)))
+        y2 = self.cv2(inputs)
+        return self.cv4(self.act(self.bn(tf.concat((y1, y2), axis=3))))
+
+
+class TFC3(keras.layers.Layer):
+    # CSP Bottleneck with 3 convolutions
+    def __init__(self, c1, c2, n=1, shortcut=True, g=1, e=0.5, w=None):
+        # ch_in, ch_out, number, shortcut, groups, expansion
+        super().__init__()
+        c_ = int(c2 * e)  # hidden channels
+        self.cv1 = TFConv(c1, c_, 1, 1, w=w.cv1)
+        self.cv2 = TFConv(c1, c_, 1, 1, w=w.cv2)
+        self.cv3 = TFConv(2 * c_, c2, 1, 1, w=w.cv3)
+        self.m = keras.Sequential([TFBottleneck(c_, c_, shortcut, g, e=1.0, w=w.m[j]) for j in range(n)])
+
+    def call(self, inputs):
+        return self.cv3(tf.concat((self.m(self.cv1(inputs)), self.cv2(inputs)), axis=3))
+
+
+class TFSPP(keras.layers.Layer):
+    # Spatial pyramid pooling layer used in YOLOv3-SPP
+    def __init__(self, c1, c2, k=(5, 9, 13), w=None):
+        super().__init__()
+        c_ = c1 // 2  # hidden channels
+        self.cv1 = TFConv(c1, c_, 1, 1, w=w.cv1)
+        self.cv2 = TFConv(c_ * (len(k) + 1), c2, 1, 1, w=w.cv2)
+        self.m = [keras.layers.MaxPool2D(pool_size=x, strides=1, padding='SAME') for x in k]
+
+    def call(self, inputs):
+        x = self.cv1(inputs)
+        return self.cv2(tf.concat([x] + [m(x) for m in self.m], 3))
+
+
+class TFSPPF(keras.layers.Layer):
+    # Spatial pyramid pooling-Fast layer
+    def __init__(self, c1, c2, k=5, w=None):
+        super().__init__()
+        c_ = c1 // 2  # hidden channels
+        self.cv1 = TFConv(c1, c_, 1, 1, w=w.cv1)
+        self.cv2 = TFConv(c_ * 4, c2, 1, 1, w=w.cv2)
+        self.m = keras.layers.MaxPool2D(pool_size=k, strides=1, padding='SAME')
+
+    def call(self, inputs):
+        x = self.cv1(inputs)
+        y1 = self.m(x)
+        y2 = self.m(y1)
+        return self.cv2(tf.concat([x, y1, y2, self.m(y2)], 3))
+
+
+class TFDetect(keras.layers.Layer):
+    def __init__(self, nc=80, anchors=(), ch=(), imgsz=(640, 640), w=None):  # detection layer
+        super().__init__()
+        self.stride = tf.convert_to_tensor(w.stride.numpy(), dtype=tf.float32)
+        self.nc = nc  # number of classes
+        self.no = nc + 5  # number of outputs per anchor
+        self.nl = len(anchors)  # number of detection layers
+        self.na = len(anchors[0]) // 2  # number of anchors
+        self.grid = [tf.zeros(1)] * self.nl  # init grid
+        self.anchors = tf.convert_to_tensor(w.anchors.numpy(), dtype=tf.float32)
+        self.anchor_grid = tf.reshape(self.anchors * tf.reshape(self.stride, [self.nl, 1, 1]),
+                                      [self.nl, 1, -1, 1, 2])
+        self.m = [TFConv2d(x, self.no * self.na, 1, w=w.m[i]) for i, x in enumerate(ch)]
+        self.training = False  # set to False after building model
+        self.imgsz = imgsz
+        for i in range(self.nl):
+            ny, nx = self.imgsz[0] // self.stride[i], self.imgsz[1] // self.stride[i]
+            self.grid[i] = self._make_grid(nx, ny)
+
+    def call(self, inputs):
+        z = []  # inference output
+        x = []
+        for i in range(self.nl):
+            x.append(self.m[i](inputs[i]))
+            # x(bs,20,20,255) to x(bs,3,20,20,85)
+            ny, nx = self.imgsz[0] // self.stride[i], self.imgsz[1] // self.stride[i]
+            x[i] = tf.transpose(tf.reshape(x[i], [-1, ny * nx, self.na, self.no]), [0, 2, 1, 3])
+
+            if not self.training:  # inference
+                y = tf.sigmoid(x[i])
+                xy = (y[..., 0:2] * 2 - 0.5 + self.grid[i]) * self.stride[i]  # xy
+                wh = (y[..., 2:4] * 2) ** 2 * self.anchor_grid[i]
+                # Normalize xywh to 0-1 to reduce calibration error
+                xy /= tf.constant([[self.imgsz[1], self.imgsz[0]]], dtype=tf.float32)
+                wh /= tf.constant([[self.imgsz[1], self.imgsz[0]]], dtype=tf.float32)
+                y = tf.concat([xy, wh, y[..., 4:]], -1)
+                z.append(tf.reshape(y, [-1, self.na * ny * nx, self.no]))
+
+        return x if self.training else (tf.concat(z, 1), x)
+
+    @staticmethod
+    def _make_grid(nx=20, ny=20):
+        # yv, xv = torch.meshgrid([torch.arange(ny), torch.arange(nx)])
+        # return torch.stack((xv, yv), 2).view((1, 1, ny, nx, 2)).float()
+        xv, yv = tf.meshgrid(tf.range(nx), tf.range(ny))
+        return tf.cast(tf.reshape(tf.stack([xv, yv], 2), [1, 1, ny * nx, 2]), dtype=tf.float32)
+
+
+class TFUpsample(keras.layers.Layer):
+    def __init__(self, size, scale_factor, mode, w=None):  # warning: all arguments needed including 'w'
+        super().__init__()
+        assert scale_factor == 2, "scale_factor must be 2"
+        self.upsample = lambda x: tf.image.resize(x, (x.shape[1] * 2, x.shape[2] * 2), method=mode)
+        # self.upsample = keras.layers.UpSampling2D(size=scale_factor, interpolation=mode)
+        # with default arguments: align_corners=False, half_pixel_centers=False
+        # self.upsample = lambda x: tf.raw_ops.ResizeNearestNeighbor(images=x,
+        #                                                            size=(x.shape[1] * 2, x.shape[2] * 2))
+
+    def call(self, inputs):
+        return self.upsample(inputs)
+
+
+class TFConcat(keras.layers.Layer):
+    def __init__(self, dimension=1, w=None):
+        super().__init__()
+        assert dimension == 1, "convert only NCHW to NHWC concat"
+        self.d = 3
+
+    def call(self, inputs):
+        return tf.concat(inputs, self.d)
+
+
+def parse_model(d, ch, model, imgsz):  # model_dict, input_channels(3)
+    LOGGER.info(f"\n{'':>3}{'from':>18}{'n':>3}{'params':>10}  {'module':<40}{'arguments':<30}")
+    anchors, nc, gd, gw = d['anchors'], d['nc'], d['depth_multiple'], d['width_multiple']
+    na = (len(anchors[0]) // 2) if isinstance(anchors, list) else anchors  # number of anchors
+    no = na * (nc + 5)  # number of outputs = anchors * (classes + 5)
+
+    layers, save, c2 = [], [], ch[-1]  # layers, savelist, ch out
+    for i, (f, n, m, args) in enumerate(d['backbone'] + d['head']):  # from, number, module, args
+        m_str = m
+        m = eval(m) if isinstance(m, str) else m  # eval strings
+        for j, a in enumerate(args):
+            try:
+                args[j] = eval(a) if isinstance(a, str) else a  # eval strings
+            except NameError:
+                pass
+
+        n = max(round(n * gd), 1) if n > 1 else n  # depth gain
+        if m in [nn.Conv2d, Conv, Bottleneck, SPP, SPPF, DWConv, MixConv2d, Focus, CrossConv, BottleneckCSP, C3]:
+            c1, c2 = ch[f], args[0]
+            c2 = make_divisible(c2 * gw, 8) if c2 != no else c2
+
+            args = [c1, c2, *args[1:]]
+            if m in [BottleneckCSP, C3]:
+                args.insert(2, n)
+                n = 1
+        elif m is nn.BatchNorm2d:
+            args = [ch[f]]
+        elif m is Concat:
+            c2 = sum(ch[-1 if x == -1 else x + 1] for x in f)
+        elif m is Detect:
+            args.append([ch[x + 1] for x in f])
+            if isinstance(args[1], int):  # number of anchors
+                args[1] = [list(range(args[1] * 2))] * len(f)
+            args.append(imgsz)
+        else:
+            c2 = ch[f]
+
+        tf_m = eval('TF' + m_str.replace('nn.', ''))
+        m_ = keras.Sequential([tf_m(*args, w=model.model[i][j]) for j in range(n)]) if n > 1 \
+            else tf_m(*args, w=model.model[i])  # module
+
+        torch_m_ = nn.Sequential(*(m(*args) for _ in range(n))) if n > 1 else m(*args)  # module
+        t = str(m)[8:-2].replace('__main__.', '')  # module type
+        np = sum(x.numel() for x in torch_m_.parameters())  # number params
+        m_.i, m_.f, m_.type, m_.np = i, f, t, np  # attach index, 'from' index, type, number params
+        LOGGER.info(f'{i:>3}{str(f):>18}{str(n):>3}{np:>10}  {t:<40}{str(args):<30}')  # print
+        save.extend(x % i for x in ([f] if isinstance(f, int) else f) if x != -1)  # append to savelist
+        layers.append(m_)
+        ch.append(c2)
+    return keras.Sequential(layers), sorted(save)
+
+
+class TFModel:
+    def __init__(self, cfg='yolov5s.yaml', ch=3, nc=None, model=None, imgsz=(640, 640)):  # model, channels, classes
+        super().__init__()
+        if isinstance(cfg, dict):
+            self.yaml = cfg  # model dict
+        else:  # is *.yaml
+            import yaml  # for torch hub
+            self.yaml_file = Path(cfg).name
+            with open(cfg) as f:
+                self.yaml = yaml.load(f, Loader=yaml.FullLoader)  # model dict
+
+        # Define model
+        if nc and nc != self.yaml['nc']:
+            LOGGER.info(f"Overriding {cfg} nc={self.yaml['nc']} with nc={nc}")
+            self.yaml['nc'] = nc  # override yaml value
+        self.model, self.savelist = parse_model(deepcopy(self.yaml), ch=[ch], model=model, imgsz=imgsz)
+
+    def predict(self, inputs, tf_nms=False, agnostic_nms=False, topk_per_class=100, topk_all=100, iou_thres=0.45,
+                conf_thres=0.25):
+        y = []  # outputs
+        x = inputs
+        for i, m in enumerate(self.model.layers):
+            if m.f != -1:  # if not from previous layer
+                x = y[m.f] if isinstance(m.f, int) else [x if j == -1 else y[j] for j in m.f]  # from earlier layers
+
+            x = m(x)  # run
+            y.append(x if m.i in self.savelist else None)  # save output
+
+        # Add TensorFlow NMS
+        if tf_nms:
+            boxes = self._xywh2xyxy(x[0][..., :4])
+            probs = x[0][:, :, 4:5]
+            classes = x[0][:, :, 5:]
+            scores = probs * classes
+            if agnostic_nms:
+                nms = AgnosticNMS()((boxes, classes, scores), topk_all, iou_thres, conf_thres)
+                return nms, x[1]
+            else:
+                boxes = tf.expand_dims(boxes, 2)
+                nms = tf.image.combined_non_max_suppression(
+                    boxes, scores, topk_per_class, topk_all, iou_thres, conf_thres, clip_boxes=False)
+                return nms, x[1]
+
+        return x[0]  # output only first tensor [1,6300,85] = [xywh, conf, class0, class1, ...]
+        # x = x[0][0]  # [x(1,6300,85), ...] to x(6300,85)
+        # xywh = x[..., :4]  # x(6300,4) boxes
+        # conf = x[..., 4:5]  # x(6300,1) confidences
+        # cls = tf.reshape(tf.cast(tf.argmax(x[..., 5:], axis=1), tf.float32), (-1, 1))  # x(6300,1)  classes
+        # return tf.concat([conf, cls, xywh], 1)
+
+    @staticmethod
+    def _xywh2xyxy(xywh):
+        # Convert nx4 boxes from [x, y, w, h] to [x1, y1, x2, y2] where xy1=top-left, xy2=bottom-right
+        x, y, w, h = tf.split(xywh, num_or_size_splits=4, axis=-1)
+        return tf.concat([x - w / 2, y - h / 2, x + w / 2, y + h / 2], axis=-1)
+
+
+class AgnosticNMS(keras.layers.Layer):
+    # TF Agnostic NMS
+    def call(self, input, topk_all, iou_thres, conf_thres):
+        # wrap map_fn to avoid TypeSpec related error https://stackoverflow.com/a/65809989/3036450
+        return tf.map_fn(lambda x: self._nms(x, topk_all, iou_thres, conf_thres), input,
+                         fn_output_signature=(tf.float32, tf.float32, tf.float32, tf.int32),
+                         name='agnostic_nms')
+
+    @staticmethod
+    def _nms(x, topk_all=100, iou_thres=0.45, conf_thres=0.25):  # agnostic NMS
+        boxes, classes, scores = x
+        class_inds = tf.cast(tf.argmax(classes, axis=-1), tf.float32)
+        scores_inp = tf.reduce_max(scores, -1)
+        selected_inds = tf.image.non_max_suppression(
+            boxes, scores_inp, max_output_size=topk_all, iou_threshold=iou_thres, score_threshold=conf_thres)
+        selected_boxes = tf.gather(boxes, selected_inds)
+        padded_boxes = tf.pad(selected_boxes,
+                              paddings=[[0, topk_all - tf.shape(selected_boxes)[0]], [0, 0]],
+                              mode="CONSTANT", constant_values=0.0)
+        selected_scores = tf.gather(scores_inp, selected_inds)
+        padded_scores = tf.pad(selected_scores,
+                               paddings=[[0, topk_all - tf.shape(selected_boxes)[0]]],
+                               mode="CONSTANT", constant_values=-1.0)
+        selected_classes = tf.gather(class_inds, selected_inds)
+        padded_classes = tf.pad(selected_classes,
+                                paddings=[[0, topk_all - tf.shape(selected_boxes)[0]]],
+                                mode="CONSTANT", constant_values=-1.0)
+        valid_detections = tf.shape(selected_inds)[0]
+        return padded_boxes, padded_scores, padded_classes, valid_detections
+
+
+def representative_dataset_gen(dataset, ncalib=100):
+    # Representative dataset generator for use with converter.representative_dataset, returns a generator of np arrays
+    for n, (path, img, im0s, vid_cap, string) in enumerate(dataset):
+        input = np.transpose(img, [1, 2, 0])
+        input = np.expand_dims(input, axis=0).astype(np.float32)
+        input /= 255
+        yield [input]
+        if n >= ncalib:
+            break
+
+
+def run(weights=ROOT / 'yolov5s.pt',  # weights path
+        imgsz=(640, 640),  # inference size h,w
+        batch_size=1,  # batch size
+        dynamic=False,  # dynamic batch size
+        ):
+    # PyTorch model
+    im = torch.zeros((batch_size, 3, *imgsz))  # BCHW image
+    model = attempt_load(weights, map_location=torch.device('cpu'), inplace=True, fuse=False)
+    _ = model(im)  # inference
+    model.info()
+
+    # TensorFlow model
+    im = tf.zeros((batch_size, *imgsz, 3))  # BHWC image
+    tf_model = TFModel(cfg=model.yaml, model=model, nc=model.nc, imgsz=imgsz)
+    _ = tf_model.predict(im)  # inference
+
+    # Keras model
+    im = keras.Input(shape=(*imgsz, 3), batch_size=None if dynamic else batch_size)
+    keras_model = keras.Model(inputs=im, outputs=tf_model.predict(im))
+    keras_model.summary()
+
+    LOGGER.info('PyTorch, TensorFlow and Keras models successfully verified.\nUse export.py for TF model export.')
+
+
+def parse_opt():
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--weights', type=str, default=ROOT / 'yolov5s.pt', help='weights path')
+    parser.add_argument('--imgsz', '--img', '--img-size', nargs='+', type=int, default=[640], help='inference size h,w')
+    parser.add_argument('--batch-size', type=int, default=1, help='batch size')
+    parser.add_argument('--dynamic', action='store_true', help='dynamic batch size')
+    opt = parser.parse_args()
+    opt.imgsz *= 2 if len(opt.imgsz) == 1 else 1  # expand
+    print_args(FILE.stem, opt)
+    return opt
+
+
+def main(opt):
+    run(**vars(opt))
+
+
+if __name__ == "__main__":
+    opt = parse_opt()
+    main(opt)

models/yolo.py

@@ -0,0 +1,596 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+"""
+YOLO-specific modules
+
+Usage:
+    $ python path/to/models/yolo.py --cfg yolov5s.yaml
+"""
+
+import argparse
+import sys
+from copy import deepcopy
+from pathlib import Path
+
+FILE = Path(__file__).resolve()
+ROOT = FILE.parents[1]  # YOLOv5 root directory
+if str(ROOT) not in sys.path:
+    sys.path.append(str(ROOT))  # add ROOT to PATH
+# ROOT = ROOT.relative_to(Path.cwd())  # relative
+
+from models.common import *
+from models.pruned_common import *
+from models.experimental import *
+from utils.autoanchor import check_anchor_order
+from utils.general import LOGGER, check_version, check_yaml, make_divisible, print_args
+from utils.plots import feature_visualization
+from utils.torch_utils import fuse_conv_and_bn, initialize_weights, model_info, scale_img, select_device, time_sync
+
+try:
+    import thop  # for FLOPs computation
+except ImportError:
+    thop = None
+
+
+class Detect(nn.Module):
+    stride = None  # strides computed during build
+    onnx_dynamic = False  # ONNX export parameter
+
+    def __init__(self, nc=80, anchors=(), ch=(), inplace=True):  # detection layer
+        super().__init__()
+        self.nc = nc  # number of classes
+        self.no = nc + 5  # number of outputs per anchor
+        self.nl = len(anchors)  # number of detection layers
+        self.na = len(anchors[0]) // 2  # number of anchors
+        self.grid = [torch.zeros(1)] * self.nl  # init grid
+        self.anchor_grid = [torch.zeros(1)] * self.nl  # init anchor grid
+        self.register_buffer('anchors', torch.tensor(anchors).float().view(self.nl, -1, 2))  # shape(nl,na,2)
+        self.m = nn.ModuleList(nn.Conv2d(x, self.no * self.na, 1) for x in ch)  # output conv
+        self.inplace = inplace  # use in-place ops (e.g. slice assignment)
+
+    def forward(self, x):
+        z = []  # inference output
+        for i in range(self.nl):
+            x[i] = self.m[i](x[i])  # conv
+            bs, _, ny, nx = x[i].shape  # x(bs,255,20,20) to x(bs,3,20,20,85)
+            x[i] = x[i].view(bs, self.na, self.no, ny, nx).permute(0, 1, 3, 4, 2).contiguous()
+
+            if not self.training:  # inference
+                if self.onnx_dynamic or self.grid[i].shape[2:4] != x[i].shape[2:4]:
+                    self.grid[i], self.anchor_grid[i] = self._make_grid(nx, ny, i)
+
+                y = x[i].sigmoid()
+                if self.inplace:
+                    y[..., 0:2] = (y[..., 0:2] * 2 - 0.5 + self.grid[i]) * self.stride[i]  # xy
+                    y[..., 2:4] = (y[..., 2:4] * 2) ** 2 * self.anchor_grid[i]  # wh
+                else:  # for YOLOv5 on AWS Inferentia https://github.com/ultralytics/yolov5/pull/2953
+                    xy = (y[..., 0:2] * 2 - 0.5 + self.grid[i]) * self.stride[i]  # xy
+                    wh = (y[..., 2:4] * 2) ** 2 * self.anchor_grid[i]  # wh
+                    y = torch.cat((xy, wh, y[..., 4:]), -1)
+                z.append(y.view(bs, -1, self.no))
+
+        return x if self.training else (torch.cat(z, 1), x)
+
+    def _make_grid(self, nx=20, ny=20, i=0):
+        d = self.anchors[i].device
+        if check_version(torch.__version__, '1.10.0'):  # torch>=1.10.0 meshgrid workaround for torch>=0.7 compatibility
+            yv, xv = torch.meshgrid([torch.arange(ny, device=d), torch.arange(nx, device=d)], indexing='ij')
+        else:
+            yv, xv = torch.meshgrid([torch.arange(ny, device=d), torch.arange(nx, device=d)])
+        grid = torch.stack((xv, yv), 2).expand((1, self.na, ny, nx, 2)).float()
+        anchor_grid = (self.anchors[i].clone() * self.stride[i]) \
+            .view((1, self.na, 1, 1, 2)).expand((1, self.na, ny, nx, 2)).float()
+        return grid, anchor_grid
+
+
+class Model(nn.Module):
+    def __init__(self, cfg='yolov5s.yaml', ch=3, nc=None, anchors=None):  # model, input channels, number of classes
+        super().__init__()
+        if isinstance(cfg, dict):
+            self.yaml = cfg  # model dict
+        else:  # is *.yaml
+            import yaml  # for torch hub
+            self.yaml_file = Path(cfg).name
+            with open(cfg, encoding='ascii', errors='ignore') as f:
+                self.yaml = yaml.safe_load(f)  # model dict
+
+        # Define model
+        ch = self.yaml['ch'] = self.yaml.get('ch', ch)  # input channels
+        if nc and nc != self.yaml['nc']:
+            LOGGER.info(f"Overriding model.yaml nc={self.yaml['nc']} with nc={nc}")
+            self.yaml['nc'] = nc  # override yaml value
+        if anchors:
+            LOGGER.info(f'Overriding model.yaml anchors with anchors={anchors}')
+            self.yaml['anchors'] = round(anchors)  # override yaml value
+        self.model, self.save = parse_model(deepcopy(self.yaml), ch=[ch])  # model, savelist
+        self.names = [str(i) for i in range(self.yaml['nc'])]  # default names
+        self.inplace = self.yaml.get('inplace', True)
+
+        # Build strides, anchors
+        m = self.model[-1]  # Detect()
+        if isinstance(m, Detect):
+            s = 256  # 2x min stride
+            m.inplace = self.inplace
+            m.stride = torch.tensor([s / x.shape[-2] for x in self.forward(torch.zeros(1, ch, s, s))])  # forward
+            m.anchors /= m.stride.view(-1, 1, 1)
+            check_anchor_order(m)
+            self.stride = m.stride
+            self._initialize_biases()  # only run once
+
+        # Init weights, biases
+        initialize_weights(self)
+        self.info()
+        LOGGER.info('')
+
+    def forward(self, x, augment=False, profile=False, visualize=False):
+        if augment:
+            return self._forward_augment(x)  # augmented inference, None
+        return self._forward_once(x, profile, visualize)  # single-scale inference, train
+
+    def _forward_augment(self, x):
+        img_size = x.shape[-2:]  # height, width
+        s = [1, 0.83, 0.67]  # scales
+        f = [None, 3, None]  # flips (2-ud, 3-lr)
+        y = []  # outputs
+        for si, fi in zip(s, f):
+            xi = scale_img(x.flip(fi) if fi else x, si, gs=int(self.stride.max()))
+            yi = self._forward_once(xi)[0]  # forward
+            # cv2.imwrite(f'img_{si}.jpg', 255 * xi[0].cpu().numpy().transpose((1, 2, 0))[:, :, ::-1])  # save
+            yi = self._descale_pred(yi, fi, si, img_size)
+            y.append(yi)
+        y = self._clip_augmented(y)  # clip augmented tails
+        return torch.cat(y, 1), None  # augmented inference, train
+
+    def _forward_once(self, x, profile=False, visualize=False):
+        y, dt = [], []  # outputs
+        for m in self.model:
+            if m.f != -1:  # if not from previous layer
+                x = y[m.f] if isinstance(m.f, int) else [x if j == -1 else y[j] for j in m.f]  # from earlier layers
+            if profile:
+                self._profile_one_layer(m, x, dt)
+            x = m(x)  # run
+            y.append(x if m.i in self.save else None)  # save output
+            if visualize:
+                feature_visualization(x, m.type, m.i, save_dir=visualize)
+        return x
+
+    def _descale_pred(self, p, flips, scale, img_size):
+        # de-scale predictions following augmented inference (inverse operation)
+        if self.inplace:
+            p[..., :4] /= scale  # de-scale
+            if flips == 2:
+                p[..., 1] = img_size[0] - p[..., 1]  # de-flip ud
+            elif flips == 3:
+                p[..., 0] = img_size[1] - p[..., 0]  # de-flip lr
+        else:
+            x, y, wh = p[..., 0:1] / scale, p[..., 1:2] / scale, p[..., 2:4] / scale  # de-scale
+            if flips == 2:
+                y = img_size[0] - y  # de-flip ud
+            elif flips == 3:
+                x = img_size[1] - x  # de-flip lr
+            p = torch.cat((x, y, wh, p[..., 4:]), -1)
+        return p
+
+    def _clip_augmented(self, y):
+        # Clip YOLOv5 augmented inference tails
+        nl = self.model[-1].nl  # number of detection layers (P3-P5)
+        g = sum(4 ** x for x in range(nl))  # grid points
+        e = 1  # exclude layer count
+        i = (y[0].shape[1] // g) * sum(4 ** x for x in range(e))  # indices
+        y[0] = y[0][:, :-i]  # large
+        i = (y[-1].shape[1] // g) * sum(4 ** (nl - 1 - x) for x in range(e))  # indices
+        y[-1] = y[-1][:, i:]  # small
+        return y
+
+    def _profile_one_layer(self, m, x, dt):
+        c = isinstance(m, Detect)  # is final layer, copy input as inplace fix
+        o = thop.profile(m, inputs=(x.copy() if c else x,), verbose=False)[0] / 1E9 * 2 if thop else 0  # FLOPs
+        t = time_sync()
+        for _ in range(10):
+            m(x.copy() if c else x)
+        dt.append((time_sync() - t) * 100)
+        if m == self.model[0]:
+            LOGGER.info(f"{'time (ms)':>10s} {'GFLOPs':>10s} {'params':>10s}  {'module'}")
+        LOGGER.info(f'{dt[-1]:10.2f} {o:10.2f} {m.np:10.0f}  {m.type}')
+        if c:
+            LOGGER.info(f"{sum(dt):10.2f} {'-':>10s} {'-':>10s}  Total")
+
+    def _initialize_biases(self, cf=None):  # initialize biases into Detect(), cf is class frequency
+        # https://arxiv.org/abs/1708.02002 section 3.3
+        # cf = torch.bincount(torch.tensor(np.concatenate(dataset.labels, 0)[:, 0]).long(), minlength=nc) + 1.
+        m = self.model[-1]  # Detect() module
+        for mi, s in zip(m.m, m.stride):  # from
+            b = mi.bias.view(m.na, -1)  # conv.bias(255) to (3,85)
+            b.data[:, 4] += math.log(8 / (640 / s) ** 2)  # obj (8 objects per 640 image)
+            b.data[:, 5:] += math.log(0.6 / (m.nc - 0.999999)) if cf is None else torch.log(cf / cf.sum())  # cls
+            mi.bias = torch.nn.Parameter(b.view(-1), requires_grad=True)
+
+    def _print_biases(self):
+        m = self.model[-1]  # Detect() module
+        for mi in m.m:  # from
+            b = mi.bias.detach().view(m.na, -1).T  # conv.bias(255) to (3,85)
+            LOGGER.info(
+                ('%6g Conv2d.bias:' + '%10.3g' * 6) % (mi.weight.shape[1], *b[:5].mean(1).tolist(), b[5:].mean()))
+
+    # def _print_weights(self):
+    #     for m in self.model.modules():
+    #         if type(m) is Bottleneck:
+    #             LOGGER.info('%10.3g' % (m.w.detach().sigmoid() * 2))  # shortcut weights
+
+    def fuse(self):  # fuse model Conv2d() + BatchNorm2d() layers
+        LOGGER.info('Fusing layers... ')
+        for m in self.model.modules():
+            if isinstance(m, (Conv, DWConv)) and hasattr(m, 'bn'):
+                m.conv = fuse_conv_and_bn(m.conv, m.bn)  # update conv
+                delattr(m, 'bn')  # remove batchnorm
+                m.forward = m.forward_fuse  # update forward
+        self.info()
+        return self
+
+    def info(self, verbose=False, img_size=640):  # print model information
+        model_info(self, verbose, img_size)
+
+    def _apply(self, fn):
+        # Apply to(), cpu(), cuda(), half() to model tensors that are not parameters or registered buffers
+        self = super()._apply(fn)
+        m = self.model[-1]  # Detect()
+        if isinstance(m, Detect):
+            m.stride = fn(m.stride)
+            m.grid = list(map(fn, m.grid))
+            if isinstance(m.anchor_grid, list):
+                m.anchor_grid = list(map(fn, m.anchor_grid))
+        return self
+
+class ModelPruned(nn.Module):
+    def __init__(self, maskbndict, cfg='yolov5s.yaml', ch=3, nc=None, anchors=None):  # model, input channels, number of classes
+        super().__init__()
+        self.maskbndict = maskbndict
+        if isinstance(cfg, dict):
+            self.yaml = cfg  # model dict
+        else:  # is *.yaml
+            import yaml  # for torch hub
+            self.yaml_file = Path(cfg).name
+            with open(cfg, encoding='ascii', errors='ignore') as f:
+                self.yaml = yaml.safe_load(f)  # model dict
+
+        # Define model
+        ch = self.yaml['ch'] = self.yaml.get('ch', ch)  # input channels
+        if nc and nc != self.yaml['nc']:
+            LOGGER.info(f"Overriding model.yaml nc={self.yaml['nc']} with nc={nc}")
+            self.yaml['nc'] = nc  # override yaml value
+        if anchors:
+            LOGGER.info(f'Overriding model.yaml anchors with anchors={anchors}')
+            self.yaml['anchors'] = round(anchors)  # override yaml value
+        self.model, self.save, self.from_to_map = parse_pruned_model(self.maskbndict, deepcopy(self.yaml), ch=[ch])  # model, savelist
+        self.names = [str(i) for i in range(self.yaml['nc'])]  # default names
+        self.inplace = self.yaml.get('inplace', True)
+
+        # Build strides, anchors
+        m = self.model[-1]  # Detect()
+        if isinstance(m, Detect):
+            s = 256  # 2x min stride
+            m.inplace = self.inplace
+            m.stride = torch.tensor([s / x.shape[-2] for x in self.forward(torch.zeros(1, ch, s, s))])  # forward
+            m.anchors /= m.stride.view(-1, 1, 1)
+            check_anchor_order(m)
+            self.stride = m.stride
+            self._initialize_biases()  # only run once
+
+        # Init weights, biases
+        initialize_weights(self)
+
+    def forward(self, x, augment=False, profile=False, visualize=False):
+        if augment:
+            return self._forward_augment(x)  # augmented inference, None
+        return self._forward_once(x, profile, visualize)  # single-scale inference, train
+
+    def _forward_augment(self, x):
+        img_size = x.shape[-2:]  # height, width
+        s = [1, 0.83, 0.67]  # scales
+        f = [None, 3, None]  # flips (2-ud, 3-lr)
+        y = []  # outputs
+        for si, fi in zip(s, f):
+            xi = scale_img(x.flip(fi) if fi else x, si, gs=int(self.stride.max()))
+            yi = self._forward_once(xi)[0]  # forward
+            # cv2.imwrite(f'img_{si}.jpg', 255 * xi[0].cpu().numpy().transpose((1, 2, 0))[:, :, ::-1])  # save
+            yi = self._descale_pred(yi, fi, si, img_size)
+            y.append(yi)
+        y = self._clip_augmented(y)  # clip augmented tails
+        return torch.cat(y, 1), None  # augmented inference, train
+
+    def _forward_once(self, x, profile=False, visualize=False):
+        y, dt = [], []  # outputs
+        for m in self.model:
+            if m.f != -1:  # if not from previous layer
+                x = y[m.f] if isinstance(m.f, int) else [x if j == -1 else y[j] for j in m.f]  # from earlier layers
+            if profile:
+                self._profile_one_layer(m, x, dt)
+            x = m(x)  # run
+            y.append(x if m.i in self.save else None)  # save output
+            if visualize:
+                feature_visualization(x, m.type, m.i, save_dir=visualize)
+        return x
+
+    def _descale_pred(self, p, flips, scale, img_size):
+        # de-scale predictions following augmented inference (inverse operation)
+        if self.inplace:
+            p[..., :4] /= scale  # de-scale
+            if flips == 2:
+                p[..., 1] = img_size[0] - p[..., 1]  # de-flip ud
+            elif flips == 3:
+                p[..., 0] = img_size[1] - p[..., 0]  # de-flip lr
+        else:
+            x, y, wh = p[..., 0:1] / scale, p[..., 1:2] / scale, p[..., 2:4] / scale  # de-scale
+            if flips == 2:
+                y = img_size[0] - y  # de-flip ud
+            elif flips == 3:
+                x = img_size[1] - x  # de-flip lr
+            p = torch.cat((x, y, wh, p[..., 4:]), -1)
+        return p
+
+    def _clip_augmented(self, y):
+        # Clip YOLOv5 augmented inference tails
+        nl = self.model[-1].nl  # number of detection layers (P3-P5)
+        g = sum(4 ** x for x in range(nl))  # grid points
+        e = 1  # exclude layer count
+        i = (y[0].shape[1] // g) * sum(4 ** x for x in range(e))  # indices
+        y[0] = y[0][:, :-i]  # large
+        i = (y[-1].shape[1] // g) * sum(4 ** (nl - 1 - x) for x in range(e))  # indices
+        y[-1] = y[-1][:, i:]  # small
+        return y
+
+    def _profile_one_layer(self, m, x, dt):
+        c = isinstance(m, Detect)  # is final layer, copy input as inplace fix
+        o = thop.profile(m, inputs=(x.copy() if c else x,), verbose=False)[0] / 1E9 * 2 if thop else 0  # FLOPs
+        t = time_sync()
+        for _ in range(10):
+            m(x.copy() if c else x)
+        dt.append((time_sync() - t) * 100)
+        if m == self.model[0]:
+            LOGGER.info(f"{'time (ms)':>10s} {'GFLOPs':>10s} {'params':>10s}  {'module'}")
+        LOGGER.info(f'{dt[-1]:10.2f} {o:10.2f} {m.np:10.0f}  {m.type}')
+        if c:
+            LOGGER.info(f"{sum(dt):10.2f} {'-':>10s} {'-':>10s}  Total")
+
+    def _initialize_biases(self, cf=None):  # initialize biases into Detect(), cf is class frequency
+        # https://arxiv.org/abs/1708.02002 section 3.3
+        # cf = torch.bincount(torch.tensor(np.concatenate(dataset.labels, 0)[:, 0]).long(), minlength=nc) + 1.
+        m = self.model[-1]  # Detect() module
+        for mi, s in zip(m.m, m.stride):  # from
+            b = mi.bias.view(m.na, -1)  # conv.bias(255) to (3,85)
+            b.data[:, 4] += math.log(8 / (640 / s) ** 2)  # obj (8 objects per 640 image)
+            b.data[:, 5:] += math.log(0.6 / (m.nc - 0.999999)) if cf is None else torch.log(cf / cf.sum())  # cls
+            mi.bias = torch.nn.Parameter(b.view(-1), requires_grad=True)
+
+    def _print_biases(self):
+        m = self.model[-1]  # Detect() module
+        for mi in m.m:  # from
+            b = mi.bias.detach().view(m.na, -1).T  # conv.bias(255) to (3,85)
+            LOGGER.info(
+                ('%6g Conv2d.bias:' + '%10.3g' * 6) % (mi.weight.shape[1], *b[:5].mean(1).tolist(), b[5:].mean()))
+
+    # def _print_weights(self):
+    #     for m in self.model.modules():
+    #         if type(m) is Bottleneck:
+    #             LOGGER.info('%10.3g' % (m.w.detach().sigmoid() * 2))  # shortcut weights
+
+    def fuse(self):  # fuse model Conv2d() + BatchNorm2d() layers
+        LOGGER.info('Fusing layers... ')
+        for m in self.model.modules():
+            if isinstance(m, (Conv, DWConv)) and hasattr(m, 'bn'):
+                m.conv = fuse_conv_and_bn(m.conv, m.bn)  # update conv
+                delattr(m, 'bn')  # remove batchnorm
+                m.forward = m.forward_fuse  # update forward
+        self.info()
+        return self
+
+    def info(self, verbose=False, img_size=640):  # print model information
+        model_info(self, verbose, img_size)
+
+    def _apply(self, fn):
+        # Apply to(), cpu(), cuda(), half() to model tensors that are not parameters or registered buffers
+        self = super()._apply(fn)
+        m = self.model[-1]  # Detect()
+        if isinstance(m, Detect):
+            m.stride = fn(m.stride)
+            m.grid = list(map(fn, m.grid))
+            if isinstance(m.anchor_grid, list):
+                m.anchor_grid = list(map(fn, m.anchor_grid))
+        return self
+
+
+def parse_model(d, ch):  # model_dict, input_channels(3)
+    LOGGER.info(f"\n{'':>3}{'from':>18}{'n':>3}{'params':>10}  {'module':<40}{'arguments':<30}")
+    anchors, nc, gd, gw = d['anchors'], d['nc'], d['depth_multiple'], d['width_multiple']
+    na = (len(anchors[0]) // 2) if isinstance(anchors, list) else anchors  # number of anchors
+    no = na * (nc + 5)  # number of outputs = anchors * (classes + 5)
+
+    layers, save, c2 = [], [], ch[-1]  # layers, savelist, ch out
+    for i, (f, n, m, args) in enumerate(d['backbone'] + d['head']):  # from, number, module, args
+        m = eval(m) if isinstance(m, str) else m  # eval strings
+        for j, a in enumerate(args):
+            try:
+                args[j] = eval(a) if isinstance(a, str) else a  # eval strings
+            except NameError:
+                pass
+
+        n = n_ = max(round(n * gd), 1) if n > 1 else n  # depth gain
+        if m in [Conv, GhostConv, Bottleneck, GhostBottleneck, SPP, SPPF, DWConv, MixConv2d, Focus, CrossConv,
+                 BottleneckCSP, C3, C3TR, C3SPP, C3Ghost]:
+            c1, c2 = ch[f], args[0]
+            if c2 != no:  # if not output
+                c2 = make_divisible(c2 * gw, 8)
+
+            args = [c1, c2, *args[1:]]
+            if m in [BottleneckCSP, C3, C3TR, C3Ghost]:
+                args.insert(2, n)  # number of repeats
+                n = 1
+        elif m is nn.BatchNorm2d:
+            args = [ch[f]]
+        elif m is Concat:
+            c2 = sum(ch[x] for x in f)
+        elif m is Detect:
+            args.append([ch[x] for x in f])
+            if isinstance(args[1], int):  # number of anchors
+                args[1] = [list(range(args[1] * 2))] * len(f)
+        elif m is Contract:
+            c2 = ch[f] * args[0] ** 2
+        elif m is Expand:
+            c2 = ch[f] // args[0] ** 2
+        else:
+            c2 = ch[f]
+
+        m_ = nn.Sequential(*(m(*args) for _ in range(n))) if n > 1 else m(*args)  # module
+        t = str(m)[8:-2].replace('__main__.', '')  # module type
+        np = sum(x.numel() for x in m_.parameters())  # number params
+        m_.i, m_.f, m_.type, m_.np = i, f, t, np  # attach index, 'from' index, type, number params
+        LOGGER.info(f'{i:>3}{str(f):>18}{n_:>3}{np:10.0f}  {t:<40}{str(args):<30}')  # print
+        save.extend(x % i for x in ([f] if isinstance(f, int) else f) if x != -1)  # append to savelist
+        layers.append(m_)
+        if i == 0:
+            ch = []
+        ch.append(c2)
+    return nn.Sequential(*layers), sorted(save)
+
+
+def parse_pruned_model(maskbndict, d, ch):  # model_dict, input_channels(3)
+    LOGGER.info(f"\n{'':>3}{'from':>18}{'n':>3}{'params':>10}  {'module':<40}{'arguments':<30}")
+    anchors, nc, gd, gw = d['anchors'], d['nc'], d['depth_multiple'], d['width_multiple']
+    na = (len(anchors[0]) // 2) if isinstance(anchors, list) else anchors  # number of anchors
+    no = na * (nc + 5)  # number of outputs = anchors * (classes + 5)
+    ch = [3]
+    fromlayer = []  # last module bn layer name
+    from_to_map = {}
+    layers, save, c2 = [], [], ch[-1]  # layers, savelist, ch out
+    for i, (f, n, m, args) in enumerate(d['backbone'] + d['head']):  # from, number, module, args
+        m = eval(m) if isinstance(m, str) else m  # eval strings
+        for j, a in enumerate(args):
+            try:
+                args[j] = eval(a) if isinstance(a, str) else a  # eval strings
+            except NameError:
+                pass
+
+        n = n_ = max(round(n * gd), 1) if n > 1 else n  # depth gain
+        named_m_base = "model.{}".format(i)
+        if m in [Conv]:
+            named_m_bn = named_m_base + ".bn"
+
+            bnc = int(maskbndict[named_m_bn].sum())
+            c1, c2 = ch[f], bnc
+            args = [c1, c2, *args[1:]]
+            layertmp = named_m_bn
+            if i>0:
+                from_to_map[layertmp] = fromlayer[f]
+            fromlayer.append(named_m_bn)
+
+        elif m in [C3Pruned]:
+            named_m_cv1_bn = named_m_base + ".cv1.bn"
+            named_m_cv2_bn = named_m_base + ".cv2.bn"
+            named_m_cv3_bn = named_m_base + ".cv3.bn"
+            from_to_map[named_m_cv1_bn] = fromlayer[f]
+            from_to_map[named_m_cv2_bn] = fromlayer[f]
+            fromlayer.append(named_m_cv3_bn)
+
+            cv1in = ch[f]
+            cv1out = int(maskbndict[named_m_cv1_bn].sum())
+            cv2out = int(maskbndict[named_m_cv2_bn].sum())
+            cv3out = int(maskbndict[named_m_cv3_bn].sum())
+            args = [cv1in, cv1out, cv2out, cv3out, n, args[-1]]
+            bottle_args = []
+            chin = [cv1out]
+
+            c3fromlayer = [named_m_cv1_bn]
+            for p in range(n):
+                named_m_bottle_cv1_bn = named_m_base + ".m.{}.cv1.bn".format(p)
+                named_m_bottle_cv2_bn = named_m_base + ".m.{}.cv2.bn".format(p)
+                bottle_cv1in = chin[-1]
+                bottle_cv1out = int(maskbndict[named_m_bottle_cv1_bn].sum())
+                bottle_cv2out = int(maskbndict[named_m_bottle_cv2_bn].sum())
+                chin.append(bottle_cv2out)
+                bottle_args.append([bottle_cv1in, bottle_cv1out, bottle_cv2out])
+                from_to_map[named_m_bottle_cv1_bn] = c3fromlayer[p]
+                from_to_map[named_m_bottle_cv2_bn] = named_m_bottle_cv1_bn
+                c3fromlayer.append(named_m_bottle_cv2_bn)
+            args.insert(4, bottle_args)
+            c2 = cv3out
+            n = 1
+            from_to_map[named_m_cv3_bn] = [c3fromlayer[-1], named_m_cv2_bn]
+        elif m in [SPPFPruned]:
+            named_m_cv1_bn = named_m_base + ".cv1.bn"
+            named_m_cv2_bn = named_m_base + ".cv2.bn"
+            cv1in = ch[f]
+            from_to_map[named_m_cv1_bn] = fromlayer[f]
+            from_to_map[named_m_cv2_bn] = [named_m_cv1_bn]*4
+            fromlayer.append(named_m_cv2_bn)
+            cv1out = int(maskbndict[named_m_cv1_bn].sum())
+            cv2out = int(maskbndict[named_m_cv2_bn].sum())
+            args = [cv1in, cv1out, cv2out, *args[1:]]
+            c2 = cv2out
+
+        elif m is nn.BatchNorm2d:
+            args = [ch[f]]
+        elif m is Concat:
+            c2 = sum(ch[x] for x in f)
+            inputtmp = [fromlayer[x] for x in f]
+            fromlayer.append(inputtmp)
+        elif m is Detect:
+            from_to_map[named_m_base + ".m.0"] = fromlayer[f[0]]
+            from_to_map[named_m_base + ".m.1"] = fromlayer[f[1]]
+            from_to_map[named_m_base + ".m.2"] = fromlayer[f[2]]
+            args.append([ch[x] for x in f])
+            if isinstance(args[1], int):  # number of anchors
+                args[1] = [list(range(args[1] * 2))] * len(f)
+        elif m is Contract:
+            c2 = ch[f] * args[0] ** 2
+        elif m is Expand:
+            c2 = ch[f] // args[0] ** 2
+        else:
+            c2 = ch[f]
+            fromtmp = fromlayer[-1]
+            fromlayer.append(fromtmp)
+
+        m_ = nn.Sequential(*(m(*args) for _ in range(n))) if n > 1 else m(*args)  # module
+        t = str(m)[8:-2].replace('__main__.', '')  # module type
+        np = sum(x.numel() for x in m_.parameters())  # number params
+        m_.i, m_.f, m_.type, m_.np = i, f, t, np  # attach index, 'from' index, type, number params
+        LOGGER.info(f'{i:>3}{str(f):>18}{n_:>3}{np:10.0f}  {t:<40}{str(args):<30}')  # print
+        save.extend(x % i for x in ([f] if isinstance(f, int) else f) if x != -1)  # append to savelist
+        layers.append(m_)
+        if i == 0:
+            ch = []
+        ch.append(c2)
+    return nn.Sequential(*layers), sorted(save), from_to_map
+
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--cfg', type=str, default='yolov5s.yaml', help='model.yaml')
+    parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')
+    parser.add_argument('--profile', action='store_true', help='profile model speed')
+    parser.add_argument('--test', action='store_true', help='test all yolo*.yaml')
+    opt = parser.parse_args()
+    opt.cfg = check_yaml(opt.cfg)  # check YAML
+    print_args(FILE.stem, opt)
+    device = select_device(opt.device)
+
+    # Create model
+    model = Model(opt.cfg).to(device)
+    model.train()
+
+    # Profile
+    if opt.profile:
+        img = torch.rand(8 if torch.cuda.is_available() else 1, 3, 640, 640).to(device)
+        y = model(img, profile=True)
+
+    # Test all models
+    if opt.test:
+        for cfg in Path(ROOT / 'models').rglob('yolo*.yaml'):
+            try:
+                _ = Model(cfg)
+            except Exception as e:
+                print(f'Error in {cfg}: {e}')
+
+    # Tensorboard (not working https://github.com/ultralytics/yolov5/issues/2898)
+    # from torch.utils.tensorboard import SummaryWriter
+    # tb_writer = SummaryWriter('.')
+    # LOGGER.info("Run 'tensorboard --logdir=models' to view tensorboard at http://localhost:6006/")
+    # tb_writer.add_graph(torch.jit.trace(model, img, strict=False), [])  # add model graph

models/yolov5l.yaml

@@ -0,0 +1,48 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+
+# Parameters
+nc: 80  # number of classes
+depth_multiple: 1.0  # model depth multiple
+width_multiple: 1.0  # layer channel multiple
+anchors:
+  - [10,13, 16,30, 33,23]  # P3/8
+  - [30,61, 62,45, 59,119]  # P4/16
+  - [116,90, 156,198, 373,326]  # P5/32
+
+# YOLOv5 v6.0 backbone
+backbone:
+  # [from, number, module, args]
+  [[-1, 1, Conv, [64, 6, 2, 2]],  # 0-P1/2
+   [-1, 1, Conv, [128, 3, 2]],  # 1-P2/4
+   [-1, 3, C3, [128]],
+   [-1, 1, Conv, [256, 3, 2]],  # 3-P3/8
+   [-1, 6, C3, [256]],
+   [-1, 1, Conv, [512, 3, 2]],  # 5-P4/16
+   [-1, 9, C3, [512]],
+   [-1, 1, Conv, [1024, 3, 2]],  # 7-P5/32
+   [-1, 3, C3, [1024]],
+   [-1, 1, SPPF, [1024, 5]],  # 9
+  ]
+
+# YOLOv5 v6.0 head
+head:
+  [[-1, 1, Conv, [512, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 6], 1, Concat, [1]],  # cat backbone P4
+   [-1, 3, C3, [512, False]],  # 13
+
+   [-1, 1, Conv, [256, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 4], 1, Concat, [1]],  # cat backbone P3
+   [-1, 3, C3, [256, False]],  # 17 (P3/8-small)
+
+   [-1, 1, Conv, [256, 3, 2]],
+   [[-1, 14], 1, Concat, [1]],  # cat head P4
+   [-1, 3, C3, [512, False]],  # 20 (P4/16-medium)
+
+   [-1, 1, Conv, [512, 3, 2]],
+   [[-1, 10], 1, Concat, [1]],  # cat head P5
+   [-1, 3, C3, [1024, False]],  # 23 (P5/32-large)
+
+   [[17, 20, 23], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5)
+  ]

models/yolov5m.yaml

@@ -0,0 +1,48 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+
+# Parameters
+nc: 80  # number of classes
+depth_multiple: 0.67  # model depth multiple
+width_multiple: 0.75  # layer channel multiple
+anchors:
+  - [10,13, 16,30, 33,23]  # P3/8
+  - [30,61, 62,45, 59,119]  # P4/16
+  - [116,90, 156,198, 373,326]  # P5/32
+
+# YOLOv5 v6.0 backbone
+backbone:
+  # [from, number, module, args]
+  [[-1, 1, Conv, [64, 6, 2, 2]],  # 0-P1/2
+   [-1, 1, Conv, [128, 3, 2]],  # 1-P2/4
+   [-1, 3, C3, [128]],
+   [-1, 1, Conv, [256, 3, 2]],  # 3-P3/8
+   [-1, 6, C3, [256]],
+   [-1, 1, Conv, [512, 3, 2]],  # 5-P4/16
+   [-1, 9, C3, [512]],
+   [-1, 1, Conv, [1024, 3, 2]],  # 7-P5/32
+   [-1, 3, C3, [1024]],
+   [-1, 1, SPPF, [1024, 5]],  # 9
+  ]
+
+# YOLOv5 v6.0 head
+head:
+  [[-1, 1, Conv, [512, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 6], 1, Concat, [1]],  # cat backbone P4
+   [-1, 3, C3, [512, False]],  # 13
+
+   [-1, 1, Conv, [256, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 4], 1, Concat, [1]],  # cat backbone P3
+   [-1, 3, C3, [256, False]],  # 17 (P3/8-small)
+
+   [-1, 1, Conv, [256, 3, 2]],
+   [[-1, 14], 1, Concat, [1]],  # cat head P4
+   [-1, 3, C3, [512, False]],  # 20 (P4/16-medium)
+
+   [-1, 1, Conv, [512, 3, 2]],
+   [[-1, 10], 1, Concat, [1]],  # cat head P5
+   [-1, 3, C3, [1024, False]],  # 23 (P5/32-large)
+
+   [[17, 20, 23], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5)
+  ]

models/yolov5n.yaml

@@ -0,0 +1,48 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+
+# Parameters
+nc: 80  # number of classes
+depth_multiple: 0.33  # model depth multiple
+width_multiple: 0.25  # layer channel multiple
+anchors:
+  - [10,13, 16,30, 33,23]  # P3/8
+  - [30,61, 62,45, 59,119]  # P4/16
+  - [116,90, 156,198, 373,326]  # P5/32
+
+# YOLOv5 v6.0 backbone
+backbone:
+  # [from, number, module, args]
+  [[-1, 1, Conv, [64, 6, 2, 2]],  # 0-P1/2
+   [-1, 1, Conv, [128, 3, 2]],  # 1-P2/4
+   [-1, 3, C3, [128]],
+   [-1, 1, Conv, [256, 3, 2]],  # 3-P3/8
+   [-1, 6, C3, [256]],
+   [-1, 1, Conv, [512, 3, 2]],  # 5-P4/16
+   [-1, 9, C3, [512]],
+   [-1, 1, Conv, [1024, 3, 2]],  # 7-P5/32
+   [-1, 3, C3, [1024]],
+   [-1, 1, SPPF, [1024, 5]],  # 9
+  ]
+
+# YOLOv5 v6.0 head
+head:
+  [[-1, 1, Conv, [512, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 6], 1, Concat, [1]],  # cat backbone P4
+   [-1, 3, C3, [512, False]],  # 13
+
+   [-1, 1, Conv, [256, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 4], 1, Concat, [1]],  # cat backbone P3
+   [-1, 3, C3, [256, False]],  # 17 (P3/8-small)
+
+   [-1, 1, Conv, [256, 3, 2]],
+   [[-1, 14], 1, Concat, [1]],  # cat head P4
+   [-1, 3, C3, [512, False]],  # 20 (P4/16-medium)
+
+   [-1, 1, Conv, [512, 3, 2]],
+   [[-1, 10], 1, Concat, [1]],  # cat head P5
+   [-1, 3, C3, [1024, False]],  # 23 (P5/32-large)
+
+   [[17, 20, 23], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5)
+  ]

models/yolov5s.yaml

@@ -0,0 +1,48 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+
+# Parameters
+nc: 20  # number of classes
+depth_multiple: 0.33  # model depth multiple
+width_multiple: 0.50  # layer channel multiple
+anchors:
+  - [10,13, 16,30, 33,23]  # P3/8
+  - [30,61, 62,45, 59,119]  # P4/16
+  - [116,90, 156,198, 373,326]  # P5/32
+
+# YOLOv5 v6.0 backbone
+backbone:
+  # [from, number, module, args]
+  [[-1, 1, Conv, [64, 6, 2, 2]],  # 0-P1/2
+   [-1, 1, Conv, [128, 3, 2]],  # 1-P2/4
+   [-1, 3, C3, [128]],
+   [-1, 1, Conv, [256, 3, 2]],  # 3-P3/8
+   [-1, 6, C3, [256]],
+   [-1, 1, Conv, [512, 3, 2]],  # 5-P4/16
+   [-1, 9, C3, [512]],
+   [-1, 1, Conv, [1024, 3, 2]],  # 7-P5/32
+   [-1, 3, C3, [1024]],
+   [-1, 1, SPPF, [1024, 5]],  # 9
+  ]
+
+# YOLOv5 v6.0 head
+head:
+  [[-1, 1, Conv, [512, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 6], 1, Concat, [1]],  # cat backbone P4
+   [-1, 3, C3, [512, False]],  # 13
+
+   [-1, 1, Conv, [256, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 4], 1, Concat, [1]],  # cat backbone P3
+   [-1, 3, C3, [256, False]],  # 17 (P3/8-small)
+
+   [-1, 1, Conv, [256, 3, 2]],
+   [[-1, 14], 1, Concat, [1]],  # cat head P4
+   [-1, 3, C3, [512, False]],  # 20 (P4/16-medium)
+
+   [-1, 1, Conv, [512, 3, 2]],
+   [[-1, 10], 1, Concat, [1]],  # cat head P5
+   [-1, 3, C3, [1024, False]],  # 23 (P5/32-large)
+
+   [[17, 20, 23], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5)
+  ]

models/yolov5s_ball.yaml

@@ -0,0 +1,48 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+
+# Parameters
+nc: 1  # number of classes
+depth_multiple: 0.33  # model depth multiple
+width_multiple: 0.50  # layer channel multiple
+anchors:
+  - [10,13, 16,30, 33,23]  # P3/8
+  - [30,61, 62,45, 59,119]  # P4/16
+  - [116,90, 156,198, 373,326]  # P5/32
+
+# YOLOv5 v6.0 backbone
+backbone:
+  # [from, number, module, args]
+  [[-1, 1, Conv, [64, 6, 2, 2]],  # 0-P1/2
+   [-1, 1, Conv, [128, 3, 2]],  # 1-P2/4
+   [-1, 3, C3, [128]],
+   [-1, 1, Conv, [256, 3, 2]],  # 3-P3/8
+   [-1, 6, C3, [256]],
+   [-1, 1, Conv, [512, 3, 2]],  # 5-P4/16
+   [-1, 9, C3, [512]],
+   [-1, 1, Conv, [1024, 3, 2]],  # 7-P5/32
+   [-1, 3, C3, [1024]],
+   [-1, 1, SPPF, [1024, 5]],  # 9
+  ]
+
+# YOLOv5 v6.0 head
+head:
+  [[-1, 1, Conv, [512, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 6], 1, Concat, [1]],  # cat backbone P4
+   [-1, 3, C3, [512, False]],  # 13
+
+   [-1, 1, Conv, [256, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 4], 1, Concat, [1]],  # cat backbone P3
+   [-1, 3, C3, [256, False]],  # 17 (P3/8-small)
+
+   [-1, 1, Conv, [256, 3, 2]],
+   [[-1, 14], 1, Concat, [1]],  # cat head P4
+   [-1, 3, C3, [512, False]],  # 20 (P4/16-medium)
+
+   [-1, 1, Conv, [512, 3, 2]],
+   [[-1, 10], 1, Concat, [1]],  # cat head P5
+   [-1, 3, C3, [1024, False]],  # 23 (P5/32-large)
+
+   [[17, 20, 23], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5)
+  ]

models/yolov5x.yaml

@@ -0,0 +1,48 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+
+# Parameters
+nc: 80  # number of classes
+depth_multiple: 1.33  # model depth multiple
+width_multiple: 1.25  # layer channel multiple
+anchors:
+  - [10,13, 16,30, 33,23]  # P3/8
+  - [30,61, 62,45, 59,119]  # P4/16
+  - [116,90, 156,198, 373,326]  # P5/32
+
+# YOLOv5 v6.0 backbone
+backbone:
+  # [from, number, module, args]
+  [[-1, 1, Conv, [64, 6, 2, 2]],  # 0-P1/2
+   [-1, 1, Conv, [128, 3, 2]],  # 1-P2/4
+   [-1, 3, C3, [128]],
+   [-1, 1, Conv, [256, 3, 2]],  # 3-P3/8
+   [-1, 6, C3, [256]],
+   [-1, 1, Conv, [512, 3, 2]],  # 5-P4/16
+   [-1, 9, C3, [512]],
+   [-1, 1, Conv, [1024, 3, 2]],  # 7-P5/32
+   [-1, 3, C3, [1024]],
+   [-1, 1, SPPF, [1024, 5]],  # 9
+  ]
+
+# YOLOv5 v6.0 head
+head:
+  [[-1, 1, Conv, [512, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 6], 1, Concat, [1]],  # cat backbone P4
+   [-1, 3, C3, [512, False]],  # 13
+
+   [-1, 1, Conv, [256, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 4], 1, Concat, [1]],  # cat backbone P3
+   [-1, 3, C3, [256, False]],  # 17 (P3/8-small)
+
+   [-1, 1, Conv, [256, 3, 2]],
+   [[-1, 14], 1, Concat, [1]],  # cat head P4
+   [-1, 3, C3, [512, False]],  # 20 (P4/16-medium)
+
+   [-1, 1, Conv, [512, 3, 2]],
+   [[-1, 10], 1, Concat, [1]],  # cat head P5
+   [-1, 3, C3, [1024, False]],  # 23 (P5/32-large)
+
+   [[17, 20, 23], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5)
+  ]

prepare_data.py

@@ -0,0 +1,139 @@
+import xml.etree.ElementTree as ET
+import pickle
+import os
+from os import listdir, getcwd
+from os.path import join
+import random
+from shutil import copyfile
+
+classes=["ball"]
+#classes=["ball","messi"]
+
+TRAIN_RATIO = 80
+
+def clear_hidden_files(path):
+    dir_list = os.listdir(path)
+    for i in dir_list:
+        abspath = os.path.join(os.path.abspath(path), i)
+        if os.path.isfile(abspath):
+            if i.startswith("._"):
+                os.remove(abspath)
+        else:
+            clear_hidden_files(abspath)
+
+def convert(size, box):
+    dw = 1./size[0]
+    dh = 1./size[1]
+    x = (box[0] + box[1])/2.0
+    y = (box[2] + box[3])/2.0
+    w = box[1] - box[0]
+    h = box[3] - box[2]
+    x = x*dw
+    w = w*dw
+    y = y*dh
+    h = h*dh
+    return (x,y,w,h)
+
+def convert_annotation(image_id):
+    in_file = open('VOCdevkit/VOC2007/Annotations/%s.xml' %image_id)
+    out_file = open('VOCdevkit/VOC2007/YOLOLabels/%s.txt' %image_id, 'w')
+    tree=ET.parse(in_file)
+    root = tree.getroot()
+    size = root.find('size')
+    w = int(size.find('width').text)
+    h = int(size.find('height').text)
+
+    for obj in root.iter('object'):
+        difficult = obj.find('difficult').text
+        cls = obj.find('name').text
+        if cls not in classes or int(difficult) == 1:
+            continue
+        cls_id = classes.index(cls)
+        xmlbox = obj.find('bndbox')
+        b = (float(xmlbox.find('xmin').text), float(xmlbox.find('xmax').text), float(xmlbox.find('ymin').text), float(xmlbox.find('ymax').text))
+        bb = convert((w,h), b)
+        out_file.write(str(cls_id) + " " + " ".join([str(a) for a in bb]) + '\n')
+    in_file.close()
+    out_file.close()
+
+wd = os.getcwd()
+wd = os.getcwd()
+data_base_dir = os.path.join(wd, "VOCdevkit/")
+if not os.path.isdir(data_base_dir):
+    os.mkdir(data_base_dir)
+work_sapce_dir = os.path.join(data_base_dir, "VOC2007/")
+if not os.path.isdir(work_sapce_dir):
+    os.mkdir(work_sapce_dir)
+annotation_dir = os.path.join(work_sapce_dir, "Annotations/")
+if not os.path.isdir(annotation_dir):
+        os.mkdir(annotation_dir)
+clear_hidden_files(annotation_dir)
+image_dir = os.path.join(work_sapce_dir, "JPEGImages/")
+if not os.path.isdir(image_dir):
+        os.mkdir(image_dir)
+clear_hidden_files(image_dir)
+yolo_labels_dir = os.path.join(work_sapce_dir, "YOLOLabels/")
+if not os.path.isdir(yolo_labels_dir):
+        os.mkdir(yolo_labels_dir)
+clear_hidden_files(yolo_labels_dir)
+yolov5_images_dir = os.path.join(data_base_dir, "images/")
+if not os.path.isdir(yolov5_images_dir):
+        os.mkdir(yolov5_images_dir)
+clear_hidden_files(yolov5_images_dir)
+yolov5_labels_dir = os.path.join(data_base_dir, "labels/")
+if not os.path.isdir(yolov5_labels_dir):
+        os.mkdir(yolov5_labels_dir)
+clear_hidden_files(yolov5_labels_dir)
+yolov5_images_train_dir = os.path.join(yolov5_images_dir, "train/")
+if not os.path.isdir(yolov5_images_train_dir):
+        os.mkdir(yolov5_images_train_dir)
+clear_hidden_files(yolov5_images_train_dir)
+yolov5_images_test_dir = os.path.join(yolov5_images_dir, "val/")
+if not os.path.isdir(yolov5_images_test_dir):
+        os.mkdir(yolov5_images_test_dir)
+clear_hidden_files(yolov5_images_test_dir)
+yolov5_labels_train_dir = os.path.join(yolov5_labels_dir, "train/")
+if not os.path.isdir(yolov5_labels_train_dir):
+        os.mkdir(yolov5_labels_train_dir)
+clear_hidden_files(yolov5_labels_train_dir)
+yolov5_labels_test_dir = os.path.join(yolov5_labels_dir, "val/")
+if not os.path.isdir(yolov5_labels_test_dir):
+        os.mkdir(yolov5_labels_test_dir)
+clear_hidden_files(yolov5_labels_test_dir)
+
+train_file = open(os.path.join(wd, "yolov5_train.txt"), 'w')
+test_file = open(os.path.join(wd, "yolov5_val.txt"), 'w')
+train_file.close()
+test_file.close()
+train_file = open(os.path.join(wd, "yolov5_train.txt"), 'a')
+test_file = open(os.path.join(wd, "yolov5_val.txt"), 'a')
+list_imgs = os.listdir(image_dir) # list image files
+prob = random.randint(1, 100)
+print("Probability: %d" % prob)
+for i in range(0,len(list_imgs)):
+    path = os.path.join(image_dir,list_imgs[i])
+    if os.path.isfile(path):
+        image_path = image_dir + list_imgs[i]
+        voc_path = list_imgs[i]
+        (nameWithoutExtention, extention) = os.path.splitext(os.path.basename(image_path))
+        (voc_nameWithoutExtention, voc_extention) = os.path.splitext(os.path.basename(voc_path))
+        annotation_name = nameWithoutExtention + '.xml'
+        annotation_path = os.path.join(annotation_dir, annotation_name)
+        label_name = nameWithoutExtention + '.txt'
+        label_path = os.path.join(yolo_labels_dir, label_name)
+    prob = random.randint(1, 100)
+    print("Probability: %d" % prob)
+    if(prob < TRAIN_RATIO): # train dataset
+        if os.path.exists(annotation_path):
+            train_file.write(image_path + '\n')
+            convert_annotation(nameWithoutExtention) # convert label
+            copyfile(image_path, yolov5_images_train_dir + voc_path)
+            copyfile(label_path, yolov5_labels_train_dir + label_name)
+    else: # test dataset
+        if os.path.exists(annotation_path):
+            test_file.write(image_path + '\n')
+            convert_annotation(nameWithoutExtention) # convert label
+            copyfile(image_path, yolov5_images_test_dir + voc_path)
+            copyfile(label_path, yolov5_labels_test_dir + label_name)
+train_file.close()
+test_file.close()

prune.py

@@ -0,0 +1,800 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+"""
+Validate a trained YOLOv5 model accuracy on a custom dataset
+
+Usage:
+    $ python path/to/val.py --data coco128.yaml --weights yolov5s.pt --img 640
+"""
+
+import argparse
+import json
+import os
+import sys
+from pathlib import Path
+from threading import Thread
+from models.common import Bottleneck
+import numpy as np
+import torch
+from tqdm import tqdm
+import yaml  
+from utils.prune_utils import gather_bn_weights, obtain_bn_mask
+FILE = Path(__file__).resolve()
+ROOT = FILE.parents[0]  # YOLOv5 root directory
+if str(ROOT) not in sys.path:
+    sys.path.append(str(ROOT))  # add ROOT to PATH
+ROOT = Path(os.path.relpath(ROOT, Path.cwd()))  # relative
+from models.pruned_common import C3Pruned, SPPFPruned, BottleneckPruned
+from models.common import DetectMultiBackend
+from utils.callbacks import Callbacks
+from utils.datasets import create_dataloader
+from utils.general import (LOGGER, box_iou, check_dataset, check_img_size, check_requirements, check_yaml,
+                           coco80_to_coco91_class, colorstr, increment_path, non_max_suppression, print_args,
+                           scale_coords, xywh2xyxy, xyxy2xywh)
+from utils.metrics import ConfusionMatrix, ap_per_class
+from utils.plots import output_to_target, plot_images, plot_val_study
+from utils.torch_utils import select_device, time_sync
+from models.yolo import *
+
+def save_one_txt(predn, save_conf, shape, file):
+    # Save one txt result
+    gn = torch.tensor(shape)[[1, 0, 1, 0]]  # normalization gain whwh
+    for *xyxy, conf, cls in predn.tolist():
+        xywh = (xyxy2xywh(torch.tensor(xyxy).view(1, 4)) / gn).view(-1).tolist()  # normalized xywh
+        line = (cls, *xywh, conf) if save_conf else (cls, *xywh)  # label format
+        with open(file, 'a') as f:
+            f.write(('%g ' * len(line)).rstrip() % line + '\n')
+
+
+def save_one_json(predn, jdict, path, class_map):
+    # Save one JSON result {"image_id": 42, "category_id": 18, "bbox": [258.15, 41.29, 348.26, 243.78], "score": 0.236}
+    image_id = int(path.stem) if path.stem.isnumeric() else path.stem
+    box = xyxy2xywh(predn[:, :4])  # xywh
+    box[:, :2] -= box[:, 2:] / 2  # xy center to top-left corner
+    for p, b in zip(predn.tolist(), box.tolist()):
+        jdict.append({'image_id': image_id,
+                      'category_id': class_map[int(p[5])],
+                      'bbox': [round(x, 3) for x in b],
+                      'score': round(p[4], 5)})
+
+
+def process_batch(detections, labels, iouv):
+    """
+    Return correct predictions matrix. Both sets of boxes are in (x1, y1, x2, y2) format.
+    Arguments:
+        detections (Array[N, 6]), x1, y1, x2, y2, conf, class
+        labels (Array[M, 5]), class, x1, y1, x2, y2
+    Returns:
+        correct (Array[N, 10]), for 10 IoU levels
+    """
+    correct = torch.zeros(detections.shape[0], iouv.shape[0], dtype=torch.bool, device=iouv.device)
+    iou = box_iou(labels[:, 1:], detections[:, :4])
+    x = torch.where((iou >= iouv[0]) & (labels[:, 0:1] == detections[:, 5]))  # IoU above threshold and classes match
+    if x[0].shape[0]:
+        matches = torch.cat((torch.stack(x, 1), iou[x[0], x[1]][:, None]), 1).cpu().numpy()  # [label, detection, iou]
+        if x[0].shape[0] > 1:
+            matches = matches[matches[:, 2].argsort()[::-1]]
+            matches = matches[np.unique(matches[:, 1], return_index=True)[1]]
+            # matches = matches[matches[:, 2].argsort()[::-1]]
+            matches = matches[np.unique(matches[:, 0], return_index=True)[1]]
+        matches = torch.Tensor(matches).to(iouv.device)
+        correct[matches[:, 1].long()] = matches[:, 2:3] >= iouv
+    return correct
+
+
+@torch.no_grad()
+def run(data,
+        weights=None,  # model.pt path(s)
+        cfg = 'models/yolov5s.yaml',
+        percent=0,
+        batch_size=32,  # batch size
+        imgsz=640,  # inference size (pixels)
+        conf_thres=0.001,  # confidence threshold
+        iou_thres=0.6,  # NMS IoU threshold
+        task='val',  # train, val, test, speed or study
+        device='',  # cuda device, i.e. 0 or 0,1,2,3 or cpu
+        workers=8,  # max dataloader workers (per RANK in DDP mode)
+        single_cls=False,  # treat as single-class dataset
+        augment=False,  # augmented inference
+        verbose=False,  # verbose output
+        save_txt=False,  # save results to *.txt
+        save_hybrid=False,  # save label+prediction hybrid results to *.txt
+        save_conf=False,  # save confidences in --save-txt labels
+        save_json=False,  # save a COCO-JSON results file
+        project=ROOT / 'runs/val',  # save to project/name
+        name='exp',  # save to project/name
+        exist_ok=False,  # existing project/name ok, do not increment
+        half=True,  # use FP16 half-precision inference
+        dnn=False,  # use OpenCV DNN for ONNX inference
+        model=None,
+        dataloader=None,
+        save_dir=Path(''),
+        plots=True,
+        callbacks=Callbacks(),
+        compute_loss=None,
+        ):
+    # Initialize/load model and set device
+    training = model is not None
+    if training:  # called by train.py
+        device, pt, jit, engine = next(model.parameters()).device, True, False, False  # get model device, PyTorch model
+
+        half &= device.type != 'cpu'  # half precision only supported on CUDA
+        model.half() if half else model.float()
+    else:  # called directly
+        device = select_device(device, batch_size=batch_size)
+
+        # Directories
+        save_dir = increment_path(Path(project) / name, exist_ok=exist_ok)  # increment run
+        (save_dir / 'labels' if save_txt else save_dir).mkdir(parents=True, exist_ok=True)  # make dir
+
+        # Load model
+        model = DetectMultiBackend(weights, device=device, dnn=dnn)
+        stride, pt, jit, engine = model.stride, model.pt, model.jit, model.engine
+        imgsz = check_img_size(imgsz, s=stride)  # check image size
+        half &= (pt or jit or engine) and device.type != 'cpu'  # half precision only supported by PyTorch on CUDA
+        if pt or jit:
+            model.model.half() if half else model.model.float()
+        elif engine:
+            batch_size = model.batch_size
+        else:
+            half = False
+            batch_size = 1  # export.py models default to batch-size 1
+            device = torch.device('cpu')
+            LOGGER.info(f'Forcing --batch-size 1 square inference shape(1,3,{imgsz},{imgsz}) for non-PyTorch backends')
+
+        # Data
+        data = check_dataset(data)  # check
+
+    # Configure
+    model.eval()
+    is_coco = isinstance(data.get('val'), str) and data['val'].endswith('coco/val2017.txt')  # COCO dataset
+    nc = 1 if single_cls else int(data['nc'])  # number of classes
+    iouv = torch.linspace(0.5, 0.95, 10).to(device)  # iou vector for mAP@0.5:0.95
+    niou = iouv.numel()
+
+    # Dataloader
+    if not training:
+        model.warmup(imgsz=(1, 3, imgsz, imgsz), half=half)  # warmup
+        pad = 0.0 if task == 'speed' else 0.5
+        task = task if task in ('train', 'val', 'test') else 'val'  # path to train/val/test images
+        dataloader = create_dataloader(data[task], imgsz, batch_size, stride, single_cls, pad=pad, rect=pt,
+                                       workers=workers, prefix=colorstr(f'{task}: '))[0]
+
+    seen = 0
+    confusion_matrix = ConfusionMatrix(nc=nc)
+    names = {k: v for k, v in enumerate(model.names if hasattr(model, 'names') else model.module.names)}
+    class_map = coco80_to_coco91_class() if is_coco else list(range(1000))
+    s = ('%20s' + '%11s' * 6) % ('Class', 'Images', 'Labels', 'P', 'R', 'mAP@.5', 'mAP@.5:.95')
+    dt, p, r, f1, mp, mr, map50, map = [0.0, 0.0, 0.0], 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0
+    loss = torch.zeros(3, device=device)
+    jdict, stats, ap, ap_class = [], [], [], []
+    pbar = tqdm(dataloader, desc=s, bar_format='{l_bar}{bar:10}{r_bar}{bar:-10b}')  # progress bar
+    for batch_i, (im, targets, paths, shapes) in enumerate(pbar):
+        t1 = time_sync()
+        if pt or jit or engine:
+            im = im.to(device, non_blocking=True)
+            targets = targets.to(device)
+        im = im.half() if half else im.float()  # uint8 to fp16/32
+        im /= 255  # 0 - 255 to 0.0 - 1.0
+        nb, _, height, width = im.shape  # batch size, channels, height, width
+        t2 = time_sync()
+        dt[0] += t2 - t1
+
+        # Inference
+        out, train_out = model(im) if training else model(im, augment=augment, val=True)  # inference, loss outputs
+        dt[1] += time_sync() - t2
+
+        # Loss
+        if compute_loss:
+            loss += compute_loss([x.float() for x in train_out], targets)[1]  # box, obj, cls
+
+        # NMS
+        targets[:, 2:] *= torch.Tensor([width, height, width, height]).to(device)  # to pixels
+        lb = [targets[targets[:, 0] == i, 1:] for i in range(nb)] if save_hybrid else []  # for autolabelling
+        t3 = time_sync()
+        out = non_max_suppression(out, conf_thres, iou_thres, labels=lb, multi_label=True, agnostic=single_cls)
+        dt[2] += time_sync() - t3
+
+        # Metrics
+        for si, pred in enumerate(out):
+            labels = targets[targets[:, 0] == si, 1:]
+            nl = len(labels)
+            tcls = labels[:, 0].tolist() if nl else []  # target class
+            path, shape = Path(paths[si]), shapes[si][0]
+            seen += 1
+
+            if len(pred) == 0:
+                if nl:
+                    stats.append((torch.zeros(0, niou, dtype=torch.bool), torch.Tensor(), torch.Tensor(), tcls))
+                continue
+
+            # Predictions
+            if single_cls:
+                pred[:, 5] = 0
+            predn = pred.clone()
+            scale_coords(im[si].shape[1:], predn[:, :4], shape, shapes[si][1])  # native-space pred
+
+            # Evaluate
+            if nl:
+                tbox = xywh2xyxy(labels[:, 1:5])  # target boxes
+                scale_coords(im[si].shape[1:], tbox, shape, shapes[si][1])  # native-space labels
+                labelsn = torch.cat((labels[:, 0:1], tbox), 1)  # native-space labels
+                correct = process_batch(predn, labelsn, iouv)
+                if plots:
+                    confusion_matrix.process_batch(predn, labelsn)
+            else:
+                correct = torch.zeros(pred.shape[0], niou, dtype=torch.bool)
+            stats.append((correct.cpu(), pred[:, 4].cpu(), pred[:, 5].cpu(), tcls))  # (correct, conf, pcls, tcls)
+
+            # Save/log
+            if save_txt:
+                save_one_txt(predn, save_conf, shape, file=save_dir / 'labels' / (path.stem + '.txt'))
+            if save_json:
+                save_one_json(predn, jdict, path, class_map)  # append to COCO-JSON dictionary
+            callbacks.run('on_val_image_end', pred, predn, path, names, im[si])
+
+        # Plot images
+        if plots and batch_i < 3:
+            f = save_dir / f'val_batch{batch_i}_labels.jpg'  # labels
+            Thread(target=plot_images, args=(im, targets, paths, f, names), daemon=True).start()
+            f = save_dir / f'val_batch{batch_i}_pred.jpg'  # predictions
+            Thread(target=plot_images, args=(im, output_to_target(out), paths, f, names), daemon=True).start()
+
+    # Compute metrics
+    stats = [np.concatenate(x, 0) for x in zip(*stats)]  # to numpy
+    if len(stats) and stats[0].any():
+        tp, fp, p, r, f1, ap, ap_class = ap_per_class(*stats, plot=plots, save_dir=save_dir, names=names)
+        ap50, ap = ap[:, 0], ap.mean(1)  # AP@0.5, AP@0.5:0.95
+        mp, mr, map50, map = p.mean(), r.mean(), ap50.mean(), ap.mean()
+        nt = np.bincount(stats[3].astype(np.int64), minlength=nc)  # number of targets per class
+    else:
+        nt = torch.zeros(1)
+
+    # Print results
+    pf = '%20s' + '%11i' * 2 + '%11.3g' * 4  # print format
+    LOGGER.info(pf % ('all', seen, nt.sum(), mp, mr, map50, map))
+
+    # Print results per class
+    if (verbose or (nc < 50 and not training)) and nc > 1 and len(stats):
+        for i, c in enumerate(ap_class):
+            LOGGER.info(pf % (names[c], seen, nt[c], p[i], r[i], ap50[i], ap[i]))
+
+    # Print speeds
+    t = tuple(x / seen * 1E3 for x in dt)  # speeds per image
+    if not training:
+        shape = (batch_size, 3, imgsz, imgsz)
+        LOGGER.info(f'Speed: %.1fms pre-process, %.1fms inference, %.1fms NMS per image at shape {shape}' % t)
+
+    # Plots
+    if plots:
+        confusion_matrix.plot(save_dir=save_dir, names=list(names.values()))
+        callbacks.run('on_val_end')
+
+    # Save JSON
+    if save_json and len(jdict):
+        w = Path(weights[0] if isinstance(weights, list) else weights).stem if weights is not None else ''  # weights
+        anno_json = str(Path(data.get('path', '../coco')) / 'annotations/instances_val2017.json')  # annotations json
+        pred_json = str(save_dir / f"{w}_predictions.json")  # predictions json
+        LOGGER.info(f'\nEvaluating pycocotools mAP... saving {pred_json}...')
+        with open(pred_json, 'w') as f:
+            json.dump(jdict, f)
+
+        try:  # https://github.com/cocodataset/cocoapi/blob/master/PythonAPI/pycocoEvalDemo.ipynb
+            check_requirements(['pycocotools'])
+            from pycocotools.coco import COCO
+            from pycocotools.cocoeval import COCOeval
+
+            anno = COCO(anno_json)  # init annotations api
+            pred = anno.loadRes(pred_json)  # init predictions api
+            eval = COCOeval(anno, pred, 'bbox')
+            if is_coco:
+                eval.params.imgIds = [int(Path(x).stem) for x in dataloader.dataset.img_files]  # image IDs to evaluate
+            eval.evaluate()
+            eval.accumulate()
+            eval.summarize()
+            map, map50 = eval.stats[:2]  # update results (mAP@0.5:0.95, mAP@0.5)
+        except Exception as e:
+            LOGGER.info(f'pycocotools unable to run: {e}')
+
+    # Return results
+    model.float()  # for training
+    if not training:
+        s = f"\n{len(list(save_dir.glob('labels/*.txt')))} labels saved to {save_dir / 'labels'}" if save_txt else ''
+        LOGGER.info(f"Results saved to {colorstr('bold', save_dir)}{s}")
+    maps = np.zeros(nc) + map
+    for i, c in enumerate(ap_class):
+        maps[c] = ap[i]
+    return (mp, mr, map50, map, *(loss.cpu() / len(dataloader)).tolist()), maps, t
+
+@torch.no_grad()
+def run_prune(data,
+        weights=None,  # model.pt path(s)
+        cfg = 'models/yolov5s.yaml',
+        percent=0,
+        batch_size=32,  # batch size
+        imgsz=640,  # inference size (pixels)
+        conf_thres=0.001,  # confidence threshold
+        iou_thres=0.6,  # NMS IoU threshold
+        task='val',  # train, val, test, speed or study
+        device='',  # cuda device, i.e. 0 or 0,1,2,3 or cpu
+        workers=8,  # max dataloader workers (per RANK in DDP mode)
+        single_cls=False,  # treat as single-class dataset
+        augment=False,  # augmented inference
+        verbose=False,  # verbose output
+        save_txt=False,  # save results to *.txt
+        save_hybrid=False,  # save label+prediction hybrid results to *.txt
+        save_conf=False,  # save confidences in --save-txt labels
+        save_json=False,  # save a COCO-JSON results file
+        project=ROOT / 'runs/val',  # save to project/name
+        name='exp',  # save to project/name
+        exist_ok=False,  # existing project/name ok, do not increment
+        half=True,  # use FP16 half-precision inference
+        dnn=False,  # use OpenCV DNN for ONNX inference
+        model=None,
+        dataloader=None,
+        save_dir=Path(''),
+        plots=True,
+        callbacks=Callbacks(),
+        compute_loss=None,
+        ):
+    # Initialize/load model and set device
+    training = model is not None
+    if training:  # called by train.py
+        device, pt, jit, engine = next(model.parameters()).device, True, False, False  # get model device, PyTorch model
+
+        half &= device.type != 'cpu'  # half precision only supported on CUDA
+        model.half() if half else model.float()
+    else:  # called directly
+        device = select_device(device, batch_size=batch_size)
+
+        # Directories
+        save_dir = increment_path(Path(project) / name, exist_ok=exist_ok)  # increment run
+        (save_dir / 'labels' if save_txt else save_dir).mkdir(parents=True, exist_ok=True)  # make dir
+
+        # Load model
+        model = DetectMultiBackend(weights, device=device, dnn=dnn)
+        stride, pt, jit, engine = model.stride, model.pt, model.jit, model.engine
+        imgsz = check_img_size(imgsz, s=stride)  # check image size
+        # half &= (pt or jit or engine) and device.type != 'cpu'  # half precision only supported by PyTorch on CUDA
+        # if pt or jit:
+        #     model.model.half() if half else model.model.float()
+        # elif engine:
+        #     batch_size = model.batch_size
+        # else:
+        #     half = False
+        #     batch_size = 1  # export.py models default to batch-size 1
+        #     device = torch.device('cpu')
+        #     LOGGER.info(f'Forcing --batch-size 1 square inference shape(1,3,{imgsz},{imgsz}) for non-PyTorch backends')
+
+        # Data
+        data = check_dataset(data)  # check
+
+    # Configure
+    model = model.model
+    # print(model)
+    model.eval()
+    # =========================================== prune model ====================================#
+    # print("model.module_list:",model.named_children())
+    model_list = {}
+    ignore_bn_list = []
+
+    for i, layer in model.named_modules():
+        # if isinstance(layer, nn.Conv2d):
+        #     print("@Conv :",i,layer)
+        if isinstance(layer, Bottleneck):
+            if layer.add:
+                ignore_bn_list.append(i.rsplit(".",2)[0]+".cv1.bn")
+                ignore_bn_list.append(i + '.cv1.bn')
+                ignore_bn_list.append(i + '.cv2.bn')
+        if isinstance(layer, torch.nn.BatchNorm2d):
+            if i not in ignore_bn_list:
+                model_list[i] = layer
+                # print(i, layer)
+            # bnw = layer.state_dict()['weight']
+    model_list = {k:v for k,v in model_list.items() if k not in ignore_bn_list}
+  #  print("prune module :",model_list.keys())
+    prune_conv_list = [layer.replace("bn", "conv") for layer in model_list.keys()]
+    # print(prune_conv_list)
+    bn_weights = gather_bn_weights(model_list)
+    sorted_bn = torch.sort(bn_weights)[0]
+    # print("model_list:",model_list)
+    # print("bn_weights:",bn_weights)
+    # 避免剪掉所有channel的最高阈值(每个BN层的gamma的最大值的最小值即为阈值上限)
+    highest_thre = []
+    for bnlayer in model_list.values():
+        highest_thre.append(bnlayer.weight.data.abs().max().item())
+    # print("highest_thre:",highest_thre)
+    highest_thre = min(highest_thre)
+    # 找到highest_thre对应的下标对应的百分比
+    percent_limit = (sorted_bn == highest_thre).nonzero()[0, 0].item() / len(bn_weights)
+
+    print(f'Suggested Gamma threshold should be less than {highest_thre:.4f}.')
+    print(f'The corresponding prune ratio is {percent_limit:.3f}.')
+    # assert opt.percent < percent_limit, f"Prune ratio should less than {percent_limit}, otherwise it may cause error!!!"
+
+    # model_copy = deepcopy(model)
+    thre_index = int(len(sorted_bn) * opt.percent)
+    thre = sorted_bn[thre_index]
+    print(f'Gamma value that less than {thre:.4f} are set to zero!')
+    print("=" * 94)
+    print(f"|\t{'layer name':<25}{'|':<10}{'origin channels':<20}{'|':<10}{'remaining channels':<20}|")
+    remain_num = 0
+    modelstate = model.state_dict()
+    # ============================== save pruned model config yaml =================================#
+    pruned_yaml = {}
+    nc = model.model[-1].nc
+    with open(cfg, encoding='ascii', errors='ignore') as f:
+        model_yamls = yaml.safe_load(f)  # model dict
+    # # Define model
+    pruned_yaml["nc"] = model.model[-1].nc
+    pruned_yaml["depth_multiple"] = model_yamls["depth_multiple"]
+    pruned_yaml["width_multiple"] = model_yamls["width_multiple"]
+    pruned_yaml["anchors"] = model_yamls["anchors"]
+    anchors = model_yamls["anchors"]
+    pruned_yaml["backbone"] = [
+        [-1, 1, Conv, [64, 6, 2, 2]],  # 0-P1/2
+        [-1, 1, Conv, [128, 3, 2]],  # 1-P2/4
+        [-1, 3, C3Pruned, [128]],
+        [-1, 1, Conv, [256, 3, 2]],  # 3-P3/8
+        [-1, 6, C3Pruned, [256]],
+        [-1, 1, Conv, [512, 3, 2]],  # 5-P4/16
+        [-1, 9, C3Pruned, [512]],
+        [-1, 1, Conv, [1024, 3, 2]],  # 7-P5/32
+        [-1, 3, C3Pruned, [1024]],
+        [-1, 1, SPPFPruned, [1024, 5]],  # 9
+    ]
+    pruned_yaml["head"] = [
+        [-1, 1, Conv, [512, 1, 1]],
+        [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+        [[-1, 6], 1, Concat, [1]],  # cat backbone P4
+        [-1, 3, C3Pruned, [512, False]],  # 13
+
+        [-1, 1, Conv, [256, 1, 1]],
+        [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+        [[-1, 4], 1, Concat, [1]],  # cat backbone P3
+        [-1, 3, C3Pruned, [256, False]],  # 17 (P3/8-small)
+
+        [-1, 1, Conv, [256, 3, 2]],
+        [[-1, 14], 1, Concat, [1]],  # cat head P4
+        [-1, 3, C3Pruned, [512, False]],  # 20 (P4/16-medium)
+
+        [-1, 1, Conv, [512, 3, 2]],
+        [[-1, 10], 1, Concat, [1]],  # cat head P5
+        [-1, 3, C3Pruned, [1024, False]],  # 23 (P5/32-large)
+
+        [[17, 20, 23], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5)
+    ]
+
+    # ============================================================================== #
+    maskbndict = {}
+    
+    for bnname, bnlayer in model.named_modules():
+        if isinstance(bnlayer, nn.BatchNorm2d):
+            bn_module = bnlayer
+            mask = obtain_bn_mask(bn_module, thre) # 获得剪枝mask
+            if bnname in ignore_bn_list:
+                mask = torch.ones(bnlayer.weight.data.size()).cuda()
+            maskbndict[bnname] = mask
+            # print("mask:",mask)
+            remain_num += int(mask.sum())
+            bn_module.weight.data.mul_(mask)
+            bn_module.bias.data.mul_(mask)
+            # print("bn_module:", bn_module.bias)
+            print(f"|\t{bnname:<25}{'|':<10}{bn_module.weight.data.size()[0]:<20}{'|':<10}{int(mask.sum()):<20}|")
+            assert int(mask.sum()) > 0, "Number of remaining channels must greater than 0! please set lower prune percent."
+    print("=" * 94)
+   # print(maskbndict.keys())
+
+    pruned_model = ModelPruned(maskbndict=maskbndict, cfg=pruned_yaml, ch=3).cuda()
+    # Compatibility updates
+    for m in pruned_model.modules():
+        if type(m) in [nn.Hardswish, nn.LeakyReLU, nn.ReLU, nn.ReLU6, nn.SiLU, Detect, Model]:
+            m.inplace = True  # pytorch 1.7.0 compatibility
+        elif type(m) is Conv:
+            m._non_persistent_buffers_set = set()  # pytorch 1.6.0 compatibility
+
+    from_to_map = pruned_model.from_to_map
+    pruned_model_state = pruned_model.state_dict()
+    assert pruned_model_state.keys() == modelstate.keys()
+    # ======================================================================================= #
+    changed_state = []
+    for ((layername, layer),(pruned_layername, pruned_layer)) in zip(model.named_modules(), pruned_model.named_modules()):
+        assert layername == pruned_layername
+        if isinstance(layer, nn.Conv2d) and not layername.startswith("model.24"):
+            convname = layername[:-4]+"bn"
+            if convname in from_to_map.keys():
+                former = from_to_map[convname]
+                if isinstance(former, str):
+                    out_idx = np.squeeze(np.argwhere(np.asarray(maskbndict[layername[:-4] + "bn"].cpu().numpy())))
+                    in_idx = np.squeeze(np.argwhere(np.asarray(maskbndict[former].cpu().numpy())))
+                    w = layer.weight.data[:, in_idx, :, :].clone()
+                    
+                    if len(w.shape) ==3:     # remain only 1 channel.
+                        w = w.unsqueeze(1)
+                    w = w[out_idx, :, :, :].clone()
+                    
+                    pruned_layer.weight.data = w.clone()
+                    changed_state.append(layername + ".weight")
+                if isinstance(former, list):
+                    orignin = [modelstate[i+".weight"].shape[0] for i in former]
+                    formerin = []
+                    for it in range(len(former)):
+                        name = former[it]
+                        tmp = [i for i in range(maskbndict[name].shape[0]) if maskbndict[name][i] == 1]
+                        if it > 0:
+                            tmp = [k + sum(orignin[:it]) for k in tmp]
+                        formerin.extend(tmp)
+                    out_idx = np.squeeze(np.argwhere(np.asarray(maskbndict[layername[:-4] + "bn"].cpu().numpy())))
+                    w = layer.weight.data[out_idx, :, :, :].clone()
+                    pruned_layer.weight.data = w[:,formerin, :, :].clone()
+                    changed_state.append(layername + ".weight")
+            else:
+                out_idx = np.squeeze(np.argwhere(np.asarray(maskbndict[layername[:-4] + "bn"].cpu().numpy())))
+                w = layer.weight.data[out_idx, :, :, :].clone()
+                assert len(w.shape) == 4
+                pruned_layer.weight.data = w.clone()
+                changed_state.append(layername + ".weight")
+
+        if isinstance(layer,nn.BatchNorm2d):
+            out_idx = np.squeeze(np.argwhere(np.asarray(maskbndict[layername].cpu().numpy())))
+            pruned_layer.weight.data = layer.weight.data[out_idx].clone()
+            pruned_layer.bias.data = layer.bias.data[out_idx].clone()
+            pruned_layer.running_mean = layer.running_mean[out_idx].clone()
+            pruned_layer.running_var = layer.running_var[out_idx].clone()
+            changed_state.append(layername + ".weight")
+            changed_state.append(layername + ".bias")
+            changed_state.append(layername + ".running_mean")
+            changed_state.append(layername + ".running_var")
+            changed_state.append(layername + ".num_batches_tracked")
+
+        if isinstance(layer, nn.Conv2d) and layername.startswith("model.24"):
+            former = from_to_map[layername]
+            in_idx = np.squeeze(np.argwhere(np.asarray(maskbndict[former].cpu().numpy())))
+            pruned_layer.weight.data = layer.weight.data[:, in_idx, :, :]
+            pruned_layer.bias.data = layer.bias.data
+            changed_state.append(layername + ".weight")
+            changed_state.append(layername + ".bias")
+
+    missing = [i for i in pruned_model_state.keys() if i not in changed_state]
+
+    pruned_model.eval()
+    pruned_model.names = model.names
+    # =============================================================================================== #
+    torch.save({"model": model}, "original_model.pt")
+    model = pruned_model
+    torch.save({"model":model}, "pruned_model.pt")
+    model.cuda().eval()
+
+
+    is_coco = isinstance(data.get('val'), str) and data['val'].endswith('coco/val2017.txt')  # COCO dataset
+    nc = 1 if single_cls else int(data['nc'])  # number of classes
+    iouv = torch.linspace(0.5, 0.95, 10).to(device)  # iou vector for mAP@0.5:0.95
+    niou = iouv.numel()
+
+    # Dataloader
+    if not training:
+        if device.type != 'cpu':
+            model(torch.zeros(1, 3, imgsz, imgsz).to(device).type_as(next(model.parameters())))  # run once
+        # model.warmup(imgsz=(1, 3, imgsz, imgsz), half=half)  # warmup
+        pad = 0.0 if task == 'speed' else 0.5
+        task = task if task in ('train', 'val', 'test') else 'val'  # path to train/val/test images
+        dataloader = create_dataloader(data[task], imgsz, batch_size, stride, single_cls, pad=pad, rect=pt,
+                                       workers=workers, prefix=colorstr(f'{task}: '))[0]
+
+    seen = 0
+    confusion_matrix = ConfusionMatrix(nc=nc)
+    names = {k: v for k, v in enumerate(model.names if hasattr(model, 'names') else model.module.names)}
+    class_map = coco80_to_coco91_class() if is_coco else list(range(1000))
+    s = ('%20s' + '%11s' * 6) % ('Class', 'Images', 'Labels', 'P', 'R', 'mAP@.5', 'mAP@.5:.95')
+    dt, p, r, f1, mp, mr, map50, map = [0.0, 0.0, 0.0], 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0
+    loss = torch.zeros(3, device=device)
+    jdict, stats, ap, ap_class = [], [], [], []
+    pbar = tqdm(dataloader, desc=s, bar_format='{l_bar}{bar:10}{r_bar}{bar:-10b}')  # progress bar
+    for batch_i, (im, targets, paths, shapes) in enumerate(pbar):
+        t1 = time_sync()
+        if pt or jit or engine:
+            im = im.to(device, non_blocking=True)
+            targets = targets.to(device)
+        im = im.half() if half else im.float()  # uint8 to fp16/32
+        im /= 255  # 0 - 255 to 0.0 - 1.0
+        nb, _, height, width = im.shape  # batch size, channels, height, width
+        t2 = time_sync()
+        dt[0] += t2 - t1
+
+        # Inference
+        out, train_out = model(im) if training else model(im, augment=augment)  # inference, loss outputs
+        dt[1] += time_sync() - t2
+
+        # Loss
+        if compute_loss:
+            loss += compute_loss([x.float() for x in train_out], targets)[1]  # box, obj, cls
+
+        # NMS
+        targets[:, 2:] *= torch.Tensor([width, height, width, height]).to(device)  # to pixels
+        lb = [targets[targets[:, 0] == i, 1:] for i in range(nb)] if save_hybrid else []  # for autolabelling
+        t3 = time_sync()
+        out = non_max_suppression(out, conf_thres, iou_thres, labels=lb, multi_label=True, agnostic=single_cls)
+        dt[2] += time_sync() - t3
+
+        # Metrics
+        for si, pred in enumerate(out):
+            labels = targets[targets[:, 0] == si, 1:]
+            nl = len(labels)
+            tcls = labels[:, 0].tolist() if nl else []  # target class
+            path, shape = Path(paths[si]), shapes[si][0]
+            seen += 1
+
+            if len(pred) == 0:
+                if nl:
+                    stats.append((torch.zeros(0, niou, dtype=torch.bool), torch.Tensor(), torch.Tensor(), tcls))
+                continue
+
+            # Predictions
+            if single_cls:
+                pred[:, 5] = 0
+            predn = pred.clone()
+            scale_coords(im[si].shape[1:], predn[:, :4], shape, shapes[si][1])  # native-space pred
+
+            # Evaluate
+            if nl:
+                tbox = xywh2xyxy(labels[:, 1:5])  # target boxes
+                scale_coords(im[si].shape[1:], tbox, shape, shapes[si][1])  # native-space labels
+                labelsn = torch.cat((labels[:, 0:1], tbox), 1)  # native-space labels
+                correct = process_batch(predn, labelsn, iouv)
+                if plots:
+                    confusion_matrix.process_batch(predn, labelsn)
+            else:
+                correct = torch.zeros(pred.shape[0], niou, dtype=torch.bool)
+            stats.append((correct.cpu(), pred[:, 4].cpu(), pred[:, 5].cpu(), tcls))  # (correct, conf, pcls, tcls)
+
+            # Save/log
+            if save_txt:
+                save_one_txt(predn, save_conf, shape, file=save_dir / 'labels' / (path.stem + '.txt'))
+            if save_json:
+                save_one_json(predn, jdict, path, class_map)  # append to COCO-JSON dictionary
+            callbacks.run('on_val_image_end', pred, predn, path, names, im[si])
+
+        # Plot images
+        if plots and batch_i < 3:
+            f = save_dir / f'val_batch{batch_i}_labels.jpg'  # labels
+            Thread(target=plot_images, args=(im, targets, paths, f, names), daemon=True).start()
+            f = save_dir / f'val_batch{batch_i}_pred.jpg'  # predictions
+            Thread(target=plot_images, args=(im, output_to_target(out), paths, f, names), daemon=True).start()
+
+    # Compute metrics
+    stats = [np.concatenate(x, 0) for x in zip(*stats)]  # to numpy
+    if len(stats) and stats[0].any():
+        tp, fp, p, r, f1, ap, ap_class = ap_per_class(*stats, plot=plots, save_dir=save_dir, names=names)
+        ap50, ap = ap[:, 0], ap.mean(1)  # AP@0.5, AP@0.5:0.95
+        mp, mr, map50, map = p.mean(), r.mean(), ap50.mean(), ap.mean()
+        nt = np.bincount(stats[3].astype(np.int64), minlength=nc)  # number of targets per class
+    else:
+        nt = torch.zeros(1)
+
+    # Print results
+    pf = '%20s' + '%11i' * 2 + '%11.3g' * 4  # print format
+    LOGGER.info(pf % ('all', seen, nt.sum(), mp, mr, map50, map))
+
+    # Print results per class
+    if (verbose or (nc < 50 and not training)) and nc > 1 and len(stats):
+        for i, c in enumerate(ap_class):
+            LOGGER.info(pf % (names[c], seen, nt[c], p[i], r[i], ap50[i], ap[i]))
+
+    # Print speeds
+    t = tuple(x / seen * 1E3 for x in dt)  # speeds per image
+    if not training:
+        shape = (batch_size, 3, imgsz, imgsz)
+        LOGGER.info(f'Speed: %.1fms pre-process, %.1fms inference, %.1fms NMS per image at shape {shape}' % t)
+
+    # Plots
+    if plots:
+        confusion_matrix.plot(save_dir=save_dir, names=list(names.values()))
+        callbacks.run('on_val_end')
+
+    # Save JSON
+    if save_json and len(jdict):
+        w = Path(weights[0] if isinstance(weights, list) else weights).stem if weights is not None else ''  # weights
+        anno_json = str(Path(data.get('path', '../coco')) / 'annotations/instances_val2017.json')  # annotations json
+        pred_json = str(save_dir / f"{w}_predictions.json")  # predictions json
+        LOGGER.info(f'\nEvaluating pycocotools mAP... saving {pred_json}...')
+        with open(pred_json, 'w') as f:
+            json.dump(jdict, f)
+
+        try:  # https://github.com/cocodataset/cocoapi/blob/master/PythonAPI/pycocoEvalDemo.ipynb
+            check_requirements(['pycocotools'])
+            from pycocotools.coco import COCO
+            from pycocotools.cocoeval import COCOeval
+
+            anno = COCO(anno_json)  # init annotations api
+            pred = anno.loadRes(pred_json)  # init predictions api
+            eval = COCOeval(anno, pred, 'bbox')
+            if is_coco:
+                eval.params.imgIds = [int(Path(x).stem) for x in dataloader.dataset.img_files]  # image IDs to evaluate
+            eval.evaluate()
+            eval.accumulate()
+            eval.summarize()
+            map, map50 = eval.stats[:2]  # update results (mAP@0.5:0.95, mAP@0.5)
+        except Exception as e:
+            LOGGER.info(f'pycocotools unable to run: {e}')
+
+    # Return results
+    model.float()  # for training
+    if not training:
+        s = f"\n{len(list(save_dir.glob('labels/*.txt')))} labels saved to {save_dir / 'labels'}" if save_txt else ''
+        LOGGER.info(f"Results saved to {colorstr('bold', save_dir)}{s}")
+    maps = np.zeros(nc) + map
+    for i, c in enumerate(ap_class):
+        maps[c] = ap[i]
+    return (mp, mr, map50, map, *(loss.cpu() / len(dataloader)).tolist()), maps, t
+
+
+def parse_opt():
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--data', type=str, default=ROOT / 'data/VOC.yaml', help='dataset.yaml path')
+    parser.add_argument('--weights', nargs='+', type=str, default=ROOT / 'VOC2007_wm/train/exp5/weights/best.pt', help='model.pt path(s)')
+    parser.add_argument('--cfg', type=str, default='models/yolov5s.yaml', help='model.yaml path')
+    parser.add_argument('--percent', type=float, default=0.15, help='prune percentage')
+    parser.add_argument('--batch-size', type=int, default=32, help='batch size')
+    parser.add_argument('--imgsz', '--img', '--img-size', type=int, default=512, help='inference size (pixels)')
+    parser.add_argument('--conf-thres', type=float, default=0.001, help='confidence threshold')
+    parser.add_argument('--iou-thres', type=float, default=0.6, help='NMS IoU threshold')
+    parser.add_argument('--task', default='val', help='train, val, test, speed or study')
+    parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')
+    parser.add_argument('--workers', type=int, default=8, help='max dataloader workers (per RANK in DDP mode)')
+    parser.add_argument('--single-cls', action='store_true', help='treat as single-class dataset')
+    parser.add_argument('--augment', action='store_true', help='augmented inference')
+    parser.add_argument('--verbose', action='store_true', help='report mAP by class')
+    parser.add_argument('--save-txt', action='store_true', help='save results to *.txt')
+    parser.add_argument('--save-hybrid', action='store_true', help='save label+prediction hybrid results to *.txt')
+    parser.add_argument('--save-conf', action='store_true', help='save confidences in --save-txt labels')
+    parser.add_argument('--save-json', action='store_true', help='save a COCO-JSON results file')
+    parser.add_argument('--project', default=ROOT / 'VOC2007_wm/prune', help='save to project/name')
+    parser.add_argument('--name', default='exp', help='save to project/name')
+    parser.add_argument('--exist-ok', action='store_true', help='existing project/name ok, do not increment')
+    parser.add_argument('--half', action='store_true', help='use FP16 half-precision inference')
+    parser.add_argument('--dnn', action='store_true', help='use OpenCV DNN for ONNX inference')
+    opt = parser.parse_args()
+    opt.data = check_yaml(opt.data)  # check YAML
+    opt.save_json |= opt.data.endswith('coco.yaml')
+    opt.save_txt |= opt.save_hybrid
+    print_args(FILE.stem, opt)
+    return opt
+
+
+def main(opt):
+    # check_requirements(requirements=ROOT / 'requirements.txt', exclude=('tensorboard', 'thop'))
+
+    if opt.task in ('train', 'val', 'test'):  # run normally
+        if opt.conf_thres > 0.001:  # https://github.com/ultralytics/yolov5/issues/1466
+            LOGGER.info(f'WARNING: confidence threshold {opt.conf_thres} >> 0.001 will produce invalid mAP values.')
+        LOGGER.info(f'test before prune ... ')
+        run(**vars(opt))
+        LOGGER.info('='*100)
+        LOGGER.info('Test after prune ... ')
+        run_prune(**vars(opt))
+
+    else:
+        weights = opt.weights if isinstance(opt.weights, list) else [opt.weights]
+        opt.half = True  # FP16 for fastest results
+        if opt.task == 'speed':  # speed benchmarks
+            # python val.py --task speed --data coco.yaml --batch 1 --weights yolov5n.pt yolov5s.pt...
+            opt.conf_thres, opt.iou_thres, opt.save_json = 0.25, 0.45, False
+            for opt.weights in weights:
+                run(**vars(opt), plots=False)
+
+        elif opt.task == 'study':  # speed vs mAP benchmarks
+            # python val.py --task study --data coco.yaml --iou 0.7 --weights yolov5n.pt yolov5s.pt...
+            for opt.weights in weights:
+                f = f'study_{Path(opt.data).stem}_{Path(opt.weights).stem}.txt'  # filename to save to
+                x, y = list(range(256, 1536 + 128, 128)), []  # x axis (image sizes), y axis
+                for opt.imgsz in x:  # img-size
+                    LOGGER.info(f'\nRunning {f} --imgsz {opt.imgsz}...')
+                    r, _, t = run(**vars(opt), plots=False)
+                    y.append(r + t)  # results and times
+                np.savetxt(f, y, fmt='%10.4g')  # save
+            os.system('zip -r study.zip study_*.txt')
+            plot_val_study(x=x)  # plot
+
+
+if __name__ == "__main__":
+    opt = parse_opt()
+    main(opt)

requirements.txt

@@ -0,0 +1,37 @@
+# pip install -r requirements.txt
+
+# Base ----------------------------------------
+matplotlib>=3.2.2
+numpy>=1.18.5
+opencv-python>=4.1.2
+Pillow>=7.1.2
+PyYAML>=5.3.1
+requests>=2.23.0
+scipy>=1.4.1
+torch>=1.7.0
+torchvision>=0.8.1
+tqdm>=4.41.0
+
+# Logging -------------------------------------
+tensorboard>=2.4.1
+# wandb
+
+# Plotting ------------------------------------
+pandas>=1.1.4
+seaborn>=0.11.0
+
+# Export --------------------------------------
+# coremltools>=4.1  # CoreML export
+# onnx>=1.9.0  # ONNX export
+# onnx-simplifier>=0.3.6  # ONNX simplifier
+# scikit-learn==0.19.2  # CoreML quantization
+# tensorflow>=2.4.1  # TFLite export
+# tensorflowjs>=3.9.0  # TF.js export
+# openvino-dev  # OpenVINO export
+
+# Extras --------------------------------------
+# albumentations>=1.0.3
+# Cython  # for pycocotools https://github.com/cocodataset/cocoapi/issues/172
+# pycocotools>=2.0  # COCO mAP
+# roboflow
+thop  # FLOPs computation

setup.cfg

@@ -0,0 +1,45 @@
+# Project-wide configuration file, can be used for package metadata and other toll configurations
+# Example usage: global configuration for PEP8 (via flake8) setting or default pytest arguments
+
+[metadata]
+license_file = LICENSE
+description-file = README.md
+
+
+[tool:pytest]
+norecursedirs =
+    .git
+    dist
+    build
+addopts =
+    --doctest-modules
+    --durations=25
+    --color=yes
+
+
+[flake8]
+max-line-length = 120
+exclude = .tox,*.egg,build,temp
+select = E,W,F
+doctests = True
+verbose = 2
+# https://pep8.readthedocs.io/en/latest/intro.html#error-codes
+format = pylint
+# see: https://www.flake8rules.com/
+ignore =
+    E731  # Do not assign a lambda expression, use a def
+    F405  # name may be undefined, or defined from star imports: module
+    E402  # module level import not at top of file
+    F401  # module imported but unused
+    W504  # line break after binary operator
+    E127  # continuation line over-indented for visual indent
+    W504  # line break after binary operator
+    E231  # missing whitespace after ‘,’, ‘;’, or ‘:’
+    E501  # line too long
+    F403  # ‘from module import *’ used; unable to detect undefined names
+
+
+[isort]
+# https://pycqa.github.io/isort/docs/configuration/options.html
+line_length = 120
+multi_line_output = 0

tool/change_dir.py

@@ -0,0 +1,27 @@
+import argparse
+
+# -------------------------------------------------------------------------------------------------------------------- #
+# 设置
+parser = argparse.ArgumentParser(description='更改yolo格式数据集train.txt和val.txt中图片的路径')
+parser.add_argument('--data_path', default=r'D:\dataset\ObjectDetection\voc', type=str, help='|数据根目录所在目录|')
+parser.add_argument('--change_dir', default=r'D:\dataset\ObjectDetection\voc', type=str, help='|将路径中目录换成change_dir|')
+args = parser.parse_args()
+args.train_txt = args.data_path + '/train.txt'
+args.val_txt = args.data_path + '/val.txt'
+args.txt_change = args.change_dir + '/image'
+
+
+# -------------------------------------------------------------------------------------------------------------------- #
+# 程序
+def change_dir(txt):
+    with open(txt, 'r')as f:
+        label = f.readlines()
+        label = [args.txt_change + _.split('image')[-1] for _ in label]
+    with open(txt, 'w')as f:
+        f.writelines(label)
+
+
+if __name__ == '__main__':
+    change_dir(args.train_txt)
+    change_dir(args.val_txt)
+    print(f'| 已更改train.txt和val.txt中的图片根路径为:{args.change_dir} |')

tool/check_image.py

@@ -0,0 +1,26 @@
+import os
+import argparse
+
+# -------------------------------------------------------------------------------------------------------------------- #
+# 设置
+parser = argparse.ArgumentParser(description='检查train.txt和val.txt中图片是否存在')
+parser.add_argument('--data_path', default=r'D:\dataset\ObjectDetection\voc', type=str, help='|图片所在目录|')
+args = parser.parse_args()
+args.train_path = args.data_path + '/train.txt'
+args.val_path = args.data_path + '/val.txt'
+
+
+# -------------------------------------------------------------------------------------------------------------------- #
+# 程序
+def check_image(txt_path):
+    with open(txt_path, 'r')as f:
+        image_path = [_.strip() for _ in f.readlines()]
+    for i in range(len(image_path)):
+        if not os.path.exists(image_path[i]):
+            print(f'| {txt_path}:不存在{image_path[i]} |')
+
+
+if __name__ == '__main__':
+    check_image(args.train_path)
+    check_image(args.val_path)
+    print(f'| 已完成{args.data_path}中train.txt和val.txt所需要的图片检擦 |')

tool/generate_txt.py

@@ -0,0 +1,57 @@
+import os
+import yaml
+
+def generate_txt_file(data_dir, subset, txt_filename):
+    subset_dir = os.path.join(data_dir, 'images', subset)
+    image_dir = os.path.join(subset_dir+'2017')
+    print(image_dir)
+    label_dir = os.path.join(data_dir, 'labels', subset + '2017')
+    
+    image_paths = []
+    for filename in os.listdir(image_dir):
+        if filename.endswith('.jpg') or filename.endswith('.png'):
+        # if filename.endswith('.txt'):
+            image_path = os.path.join(image_dir, filename)
+            image_paths.append(image_path)
+    
+    txt_path = os.path.join(data_dir, txt_filename)
+    with open(txt_path, 'w') as f:
+        for image_path in image_paths:
+            f.write(image_path + '\n')
+
+
+def generate_class_txt(coco_dir, yaml_file):
+    yaml_path = os.path.join(coco_dir, yaml_file)
+    with open(yaml_path, 'r') as f:
+        data = yaml.safe_load(f)
+    
+    class_names = data['names']
+    class_txt_path = os.path.join(coco_dir, 'class.txt')
+    with open(class_txt_path, 'w') as f:
+        for class_name in class_names:
+            f.write(class_name + '\n')
+
+def main():
+    coco_dir = '/home/yhsun/ObjectDetection-main/datasets/VOC2007'  # 替换为你的 COCO 数据集路径
+    yaml_file = 'voc.yaml'  # COCO YAML 文件名
+
+    # # 生成 train.txt
+    # generate_txt_file(coco_dir, 'train', 'train.txt')
+    # print("Processed train dataset")
+
+    # # 生成 val.txt
+    # generate_txt_file(coco_dir, 'val', 'val.txt')
+    # print("Processed val dataset")
+
+    # # 生成 test.txt
+    # generate_txt_file(coco_dir, 'test', 'test.txt')
+    # print("Processed test dataset")
+
+    # 生成 class.txt
+    generate_class_txt(coco_dir, yaml_file)
+    print("Processed class file")
+
+    print("Finished processing COCO dataset")
+
+if __name__ == "__main__":
+    main()

tool/make_txt.py

@@ -0,0 +1,24 @@
+import os
+import argparse
+
+# -------------------------------------------------------------------------------------------------------------------- #
+# 设置
+parser = argparse.ArgumentParser(description='将文件夹中的图片按比例添加到train.txt和val.txt中')
+parser.add_argument('--data_path', default=r'/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages', type=str, help='|图片所在目录|')
+parser.add_argument('--divide', default='9,1', type=str, help='|图片划分到train.txt和val.txt的比例|')
+args = parser.parse_args()
+
+# -------------------------------------------------------------------------------------------------------------------- #
+# 程序
+if __name__ == '__main__':
+    image_dir = sorted(os.listdir(args.data_path))
+    args.divide = list(map(int, args.divide.split(',')))
+    boundary = int(len(image_dir) * args.divide[0] / (args.divide[0] + args.divide[1]))
+    with open('train.txt', 'a')as f:
+        for i in range(boundary):
+            label = args.data_path + '/' + image_dir[i]
+            f.write(label + '\n')
+    with open('val.txt', 'a')as f:
+        for i in range(boundary, len(image_dir)):
+            label = args.data_path + '/' + image_dir[i]
+            f.write(label + '\n')

tool/val.txt

@@ -0,0 +1,997 @@
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008967.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008968.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008969.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008970.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008971.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008972.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008973.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008974.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008975.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008976.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008977.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008978.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008979.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008980.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008981.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008982.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008983.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008984.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008985.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008986.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008987.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008988.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008989.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008990.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008991.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008992.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008993.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008994.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008995.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008996.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008997.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008998.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/008999.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009000.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009001.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009002.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009003.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009004.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009005.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009006.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009007.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009008.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009009.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009010.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009011.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009012.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009013.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009014.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009015.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009016.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009017.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009018.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009019.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009020.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009021.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009022.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009023.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009024.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009025.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009026.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009027.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009028.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009029.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009030.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009031.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009032.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009033.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009034.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009035.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009036.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009037.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009038.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009039.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009040.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009041.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009042.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009043.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009044.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009045.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009046.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009047.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009048.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009049.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009050.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009051.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009052.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009053.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009054.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009055.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009056.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009057.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009058.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009059.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009060.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009061.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009062.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009063.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009064.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009065.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009066.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009067.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009068.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009069.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009070.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009071.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009072.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009073.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009074.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009075.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009076.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009077.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009078.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009079.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009080.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009081.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009082.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009083.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009084.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009085.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009086.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009087.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009088.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009089.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009090.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009091.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009092.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009093.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009094.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009095.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009096.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009097.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009098.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009099.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009100.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009101.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009102.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009103.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009104.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009105.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009106.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009107.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009108.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009109.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009110.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009111.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009112.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009113.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009114.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009115.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009116.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009117.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009118.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009119.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009120.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009121.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009122.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009123.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009124.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009125.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009126.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009127.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009128.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009129.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009130.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009131.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009132.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009133.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009134.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009135.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009136.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009137.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009138.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009139.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009140.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009141.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009142.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009143.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009144.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009145.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009146.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009147.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009148.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009149.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009150.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009151.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009152.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009153.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009154.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009155.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009156.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009157.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009158.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009159.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009160.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009161.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009162.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009163.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009164.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009165.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009166.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009167.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009168.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009169.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009170.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009171.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009172.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009173.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009174.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009175.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009176.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009177.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009178.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009179.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009180.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009181.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009182.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009183.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009184.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009185.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009186.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009187.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009188.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009189.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009190.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009191.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009192.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009193.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009194.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009195.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009196.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009197.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009198.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009199.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009200.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009201.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009202.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009203.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009204.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009205.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009206.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009207.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009208.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009209.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009210.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009211.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009212.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009213.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009214.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009215.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009216.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009217.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009218.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009219.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009220.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009221.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009222.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009223.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009224.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009225.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009226.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009227.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009228.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009229.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009230.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009231.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009232.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009233.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009234.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009235.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009236.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009237.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009238.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009239.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009240.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009241.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009242.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009243.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009244.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009245.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009246.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009247.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009248.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009249.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009250.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009251.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009252.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009253.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009254.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009255.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009256.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009257.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009258.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009259.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009260.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009261.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009262.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009263.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009264.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009265.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009266.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009267.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009268.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009269.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009270.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009271.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009272.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009273.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009274.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009275.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009276.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009277.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009278.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009279.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009280.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009281.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009282.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009283.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009284.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009285.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009286.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009287.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009288.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009289.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009290.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009291.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009292.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009293.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009294.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009295.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009296.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009297.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009298.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009299.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009300.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009301.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009302.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009303.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009304.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009305.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009306.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009307.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009308.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009309.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009310.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009311.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009312.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009313.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009314.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009315.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009316.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009317.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009318.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009319.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009320.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009321.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009322.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009323.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009324.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009325.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009326.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009327.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009328.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009329.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009330.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009331.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009332.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009333.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009334.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009335.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009336.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009337.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009338.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009339.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009340.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009341.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009342.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009343.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009344.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009345.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009346.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009347.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009348.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009349.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009350.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009351.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009352.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009353.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009354.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009355.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009356.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009357.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009358.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009359.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009360.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009361.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009362.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009363.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009364.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009365.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009366.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009367.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009368.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009369.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009370.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009371.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009372.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009373.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009374.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009375.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009376.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009377.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009378.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009379.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009380.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009381.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009382.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009383.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009384.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009385.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009386.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009387.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009388.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009389.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009390.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009391.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009392.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009393.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009394.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009395.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009396.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009397.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009398.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009399.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009400.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009401.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009402.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009403.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009404.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009405.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009406.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009407.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009408.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009409.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009410.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009411.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009412.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009413.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009414.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009415.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009416.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009417.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009418.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009419.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009420.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009421.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009422.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009423.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009424.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009425.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009426.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009427.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009428.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009429.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009430.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009431.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009432.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009433.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009434.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009435.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009436.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009437.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009438.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009439.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009440.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009441.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009442.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009443.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009444.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009445.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009446.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009447.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009448.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009449.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009450.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009451.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009452.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009453.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009454.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009455.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009456.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009457.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009458.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009459.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009460.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009461.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009462.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009463.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009464.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009465.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009466.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009467.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009468.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009469.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009470.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009471.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009472.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009473.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009474.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009475.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009476.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009477.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009478.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009479.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009480.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009481.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009482.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009483.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009484.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009485.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009486.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009487.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009488.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009489.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009490.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009491.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009492.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009493.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009494.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009495.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009496.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009497.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009498.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009499.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009500.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009501.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009502.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009503.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009504.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009505.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009506.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009507.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009508.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009509.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009510.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009511.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009512.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009513.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009514.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009515.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009516.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009517.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009518.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009519.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009520.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009521.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009522.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009523.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009524.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009525.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009526.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009527.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009528.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009529.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009530.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009531.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009532.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009533.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009534.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009535.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009536.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009537.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009538.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009539.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009540.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009541.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009542.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009543.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009544.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009545.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009546.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009547.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009548.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009549.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009550.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009551.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009552.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009553.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009554.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009555.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009556.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009557.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009558.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009559.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009560.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009561.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009562.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009563.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009564.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009565.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009566.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009567.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009568.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009569.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009570.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009571.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009572.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009573.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009574.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009575.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009576.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009577.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009578.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009579.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009580.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009581.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009582.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009583.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009584.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009585.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009586.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009587.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009588.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009589.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009590.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009591.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009592.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009593.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009594.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009595.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009596.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009597.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009598.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009599.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009600.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009601.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009602.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009603.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009604.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009605.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009606.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009607.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009608.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009609.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009610.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009611.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009612.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009613.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009614.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009615.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009616.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009617.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009618.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009619.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009620.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009621.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009622.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009623.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009624.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009625.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009626.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009627.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009628.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009629.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009630.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009631.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009632.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009633.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009634.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009635.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009636.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009637.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009638.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009639.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009640.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009641.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009642.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009643.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009644.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009645.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009646.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009647.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009648.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009649.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009650.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009651.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009652.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009653.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009654.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009655.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009656.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009657.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009658.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009659.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009660.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009661.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009662.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009663.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009664.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009665.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009666.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009667.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009668.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009669.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009670.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009671.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009672.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009673.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009674.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009675.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009676.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009677.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009678.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009679.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009680.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009681.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009682.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009683.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009684.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009685.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009686.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009687.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009688.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009689.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009690.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009691.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009692.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009693.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009694.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009695.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009696.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009697.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009698.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009699.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009700.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009701.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009702.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009703.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009704.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009705.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009706.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009707.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009708.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009709.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009710.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009711.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009712.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009713.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009714.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009715.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009716.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009717.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009718.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009719.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009720.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009721.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009722.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009723.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009724.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009725.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009726.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009727.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009728.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009729.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009730.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009731.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009732.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009733.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009734.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009735.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009736.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009737.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009738.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009739.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009740.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009741.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009742.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009743.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009744.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009745.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009746.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009747.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009748.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009749.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009750.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009751.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009752.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009753.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009754.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009755.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009756.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009757.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009758.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009759.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009760.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009761.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009762.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009763.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009764.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009765.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009766.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009767.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009768.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009769.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009770.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009771.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009772.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009773.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009774.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009775.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009776.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009777.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009778.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009779.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009780.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009781.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009782.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009783.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009784.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009785.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009786.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009787.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009788.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009789.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009790.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009791.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009792.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009793.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009794.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009795.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009796.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009797.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009798.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009799.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009800.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009801.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009802.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009803.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009804.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009805.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009806.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009807.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009808.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009809.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009810.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009811.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009812.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009813.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009814.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009815.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009816.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009817.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009818.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009819.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009820.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009821.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009822.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009823.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009824.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009825.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009826.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009827.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009828.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009829.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009830.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009831.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009832.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009833.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009834.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009835.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009836.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009837.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009838.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009839.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009840.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009841.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009842.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009843.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009844.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009845.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009846.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009847.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009848.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009849.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009850.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009851.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009852.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009853.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009854.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009855.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009856.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009857.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009858.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009859.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009860.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009861.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009862.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009863.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009864.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009865.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009866.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009867.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009868.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009869.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009870.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009871.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009872.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009873.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009874.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009875.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009876.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009877.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009878.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009879.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009880.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009881.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009882.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009883.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009884.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009885.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009886.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009887.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009888.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009889.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009890.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009891.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009892.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009893.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009894.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009895.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009896.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009897.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009898.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009899.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009900.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009901.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009902.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009903.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009904.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009905.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009906.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009907.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009908.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009909.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009910.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009911.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009912.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009913.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009914.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009915.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009916.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009917.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009918.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009919.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009920.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009921.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009922.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009923.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009924.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009925.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009926.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009927.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009928.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009929.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009930.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009931.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009932.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009933.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009934.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009935.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009936.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009937.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009938.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009939.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009940.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009941.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009942.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009943.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009944.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009945.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009946.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009947.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009948.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009949.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009950.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009951.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009952.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009953.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009954.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009955.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009956.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009957.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009958.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009959.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009960.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009961.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009962.jpg
+/home/yhsun/ObjectDetection-main/datasets/VOC2007/JPEGImages/009963.jpg

train.py

@@ -0,0 +1,643 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+"""
+Train a YOLOv5 model on a custom dataset.
+
+Models and datasets download automatically from the latest YOLOv5 release.
+Models: https://github.com/ultralytics/yolov5/tree/master/models
+Datasets: https://github.com/ultralytics/yolov5/tree/master/data
+Tutorial: https://github.com/ultralytics/yolov5/wiki/Train-Custom-Data
+
+Usage:
+    $ python path/to/train.py --data coco128.yaml --weights yolov5s.pt --img 640  # from pretrained (RECOMMENDED)
+    $ python path/to/train.py --data coco128.yaml --weights '' --cfg yolov5s.yaml --img 640  # from scratch
+"""
+
+import argparse
+import math
+import os
+import random
+import sys
+import time
+from copy import deepcopy
+from datetime import datetime
+from pathlib import Path
+
+import numpy as np
+import torch
+import torch.distributed as dist
+import torch.nn as nn
+import yaml
+from torch.cuda import amp
+from torch.nn.parallel import DistributedDataParallel as DDP
+from torch.optim import SGD, Adam, AdamW, lr_scheduler
+from tqdm import tqdm
+
+FILE = Path(__file__).resolve()
+ROOT = FILE.parents[0]  # YOLOv5 root directory
+if str(ROOT) not in sys.path:
+    sys.path.append(str(ROOT))  # add ROOT to PATH
+ROOT = Path(os.path.relpath(ROOT, Path.cwd()))  # relative
+
+import val  # for end-of-epoch mAP
+from models.experimental import attempt_load
+from models.yolo import Model
+from utils.autoanchor import check_anchors
+from utils.autobatch import check_train_batch_size
+from utils.callbacks import Callbacks
+from utils.datasets import create_dataloader
+from utils.downloads import attempt_download
+from utils.general import (LOGGER, check_dataset, check_file, check_git_status, check_img_size, check_requirements,
+                           check_suffix, check_yaml, colorstr, get_latest_run, increment_path, init_seeds,
+                           intersect_dicts, labels_to_class_weights, labels_to_image_weights, methods, one_cycle,
+                           print_args, print_mutation, strip_optimizer)
+from utils.loggers import Loggers
+from utils.loggers.wandb.wandb_utils import check_wandb_resume
+from utils.loss import ComputeLoss
+from utils.metrics import fitness
+from utils.plots import plot_evolve, plot_labels
+from utils.torch_utils import EarlyStopping, ModelEMA, de_parallel, select_device, torch_distributed_zero_first
+
+LOCAL_RANK = int(os.getenv('LOCAL_RANK', -1))  # https://pytorch.org/docs/stable/elastic/run.html
+RANK = int(os.getenv('RANK', -1))
+WORLD_SIZE = int(os.getenv('WORLD_SIZE', 1))
+
+
+def train(hyp,  # path/to/hyp.yaml or hyp dictionary
+          opt,
+          device,
+          callbacks
+          ):
+    save_dir, epochs, batch_size, weights, single_cls, evolve, data, cfg, resume, noval, nosave, workers, freeze = \
+        Path(opt.save_dir), opt.epochs, opt.batch_size, opt.weights, opt.single_cls, opt.evolve, opt.data, opt.cfg, \
+        opt.resume, opt.noval, opt.nosave, opt.workers, opt.freeze
+
+    # Directories
+    w = save_dir / 'weights'  # weights dir
+    (w.parent if evolve else w).mkdir(parents=True, exist_ok=True)  # make dir
+    last, best = w / 'last.pt', w / 'best.pt'
+
+    # Hyperparameters
+    if isinstance(hyp, str):
+        with open(hyp, errors='ignore') as f:
+            hyp = yaml.safe_load(f)  # load hyps dict
+    LOGGER.info(colorstr('hyperparameters: ') + ', '.join(f'{k}={v}' for k, v in hyp.items()))
+
+    # Save run settings
+    if not evolve:
+        with open(save_dir / 'hyp.yaml', 'w') as f:
+            yaml.safe_dump(hyp, f, sort_keys=False)
+        with open(save_dir / 'opt.yaml', 'w') as f:
+            yaml.safe_dump(vars(opt), f, sort_keys=False)
+
+    # Loggers
+    data_dict = None
+    if RANK in [-1, 0]:
+        loggers = Loggers(save_dir, weights, opt, hyp, LOGGER)  # loggers instance
+        if loggers.wandb:
+            data_dict = loggers.wandb.data_dict
+            if resume:
+                weights, epochs, hyp, batch_size = opt.weights, opt.epochs, opt.hyp, opt.batch_size
+
+        # Register actions
+        for k in methods(loggers):
+            callbacks.register_action(k, callback=getattr(loggers, k))
+
+    # Config
+    plots = not evolve  # create plots
+    cuda = device.type != 'cpu'
+    init_seeds(1 + RANK)
+    with torch_distributed_zero_first(LOCAL_RANK):
+        data_dict = data_dict or check_dataset(data)  # check if None
+    train_path, val_path = data_dict['train'], data_dict['val']
+    nc = 1 if single_cls else int(data_dict['nc'])  # number of classes
+    names = ['item'] if single_cls and len(data_dict['names']) != 1 else data_dict['names']  # class names
+    assert len(names) == nc, f'{len(names)} names found for nc={nc} dataset in {data}'  # check
+    is_coco = isinstance(val_path, str) and val_path.endswith('coco/val2017.txt')  # COCO dataset
+
+    # Model
+    check_suffix(weights, '.pt')  # check weights
+    pretrained = weights.endswith('.pt')
+    if pretrained:
+        with torch_distributed_zero_first(LOCAL_RANK):
+            weights = attempt_download(weights)  # download if not found locally
+        ckpt = torch.load(weights, map_location='cpu')  # load checkpoint to CPU to avoid CUDA memory leak
+        model = Model(cfg or ckpt['model'].yaml, ch=3, nc=nc, anchors=hyp.get('anchors')).to(device)  # create
+        exclude = ['anchor'] if (cfg or hyp.get('anchors')) and not resume else []  # exclude keys
+        csd = ckpt['model'].float().state_dict()  # checkpoint state_dict as FP32
+        csd = intersect_dicts(csd, model.state_dict(), exclude=exclude)  # intersect
+        model.load_state_dict(csd, strict=False)  # load
+        LOGGER.info(f'Transferred {len(csd)}/{len(model.state_dict())} items from {weights}')  # report
+    else:
+        model = Model(cfg, ch=3, nc=nc, anchors=hyp.get('anchors')).to(device)  # create
+
+    # Freeze
+    freeze = [f'model.{x}.' for x in (freeze if len(freeze) > 1 else range(freeze[0]))]  # layers to freeze
+    for k, v in model.named_parameters():
+        v.requires_grad = True  # train all layers
+        if any(x in k for x in freeze):
+            LOGGER.info(f'freezing {k}')
+            v.requires_grad = False
+
+    # Image size
+    gs = max(int(model.stride.max()), 32)  # grid size (max stride)
+    imgsz = check_img_size(opt.imgsz, gs, floor=gs * 2)  # verify imgsz is gs-multiple
+
+    # Batch size
+    if RANK == -1 and batch_size == -1:  # single-GPU only, estimate best batch size
+        batch_size = check_train_batch_size(model, imgsz)
+        loggers.on_params_update({"batch_size": batch_size})
+
+    # Optimizer
+    nbs = 64  # nominal batch size
+    accumulate = max(round(nbs / batch_size), 1)  # accumulate loss before optimizing
+    hyp['weight_decay'] *= batch_size * accumulate / nbs  # scale weight_decay
+    LOGGER.info(f"Scaled weight_decay = {hyp['weight_decay']}")
+
+    g0, g1, g2 = [], [], []  # optimizer parameter groups
+    for v in model.modules():
+        if hasattr(v, 'bias') and isinstance(v.bias, nn.Parameter):  # bias
+            g2.append(v.bias)
+        if isinstance(v, nn.BatchNorm2d):  # weight (no decay)
+            g0.append(v.weight)
+        elif hasattr(v, 'weight') and isinstance(v.weight, nn.Parameter):  # weight (with decay)
+            g1.append(v.weight)
+
+    if opt.optimizer == 'Adam':
+        optimizer = Adam(g0, lr=hyp['lr0'], betas=(hyp['momentum'], 0.999))  # adjust beta1 to momentum
+    elif opt.optimizer == 'AdamW':
+        optimizer = AdamW(g0, lr=hyp['lr0'], betas=(hyp['momentum'], 0.999))  # adjust beta1 to momentum
+    else:
+        optimizer = SGD(g0, lr=hyp['lr0'], momentum=hyp['momentum'], nesterov=True)
+
+    optimizer.add_param_group({'params': g1, 'weight_decay': hyp['weight_decay']})  # add g1 with weight_decay
+    optimizer.add_param_group({'params': g2})  # add g2 (biases)
+    LOGGER.info(f"{colorstr('optimizer:')} {type(optimizer).__name__} with parameter groups "
+                f"{len(g0)} weight (no decay), {len(g1)} weight, {len(g2)} bias")
+    del g0, g1, g2
+
+    # Scheduler
+    if opt.cos_lr:
+        lf = one_cycle(1, hyp['lrf'], epochs)  # cosine 1->hyp['lrf']
+    else:
+        lf = lambda x: (1 - x / epochs) * (1.0 - hyp['lrf']) + hyp['lrf']  # linear
+    scheduler = lr_scheduler.LambdaLR(optimizer, lr_lambda=lf)  # plot_lr_scheduler(optimizer, scheduler, epochs)
+
+    # EMA
+    ema = ModelEMA(model) if RANK in [-1, 0] else None
+
+    # Resume
+    start_epoch, best_fitness = 0, 0.0
+    if pretrained:
+        # Optimizer
+        if ckpt['optimizer'] is not None:
+            optimizer.load_state_dict(ckpt['optimizer'])
+            best_fitness = ckpt['best_fitness']
+
+        # EMA
+        if ema and ckpt.get('ema'):
+            ema.ema.load_state_dict(ckpt['ema'].float().state_dict())
+            ema.updates = ckpt['updates']
+
+        # Epochs
+        start_epoch = ckpt['epoch'] + 1
+        if resume:
+            assert start_epoch > 0, f'{weights} training to {epochs} epochs is finished, nothing to resume.'
+        if epochs < start_epoch:
+            LOGGER.info(f"{weights} has been trained for {ckpt['epoch']} epochs. Fine-tuning for {epochs} more epochs.")
+            epochs += ckpt['epoch']  # finetune additional epochs
+
+        del ckpt, csd
+
+    # DP mode
+    if cuda and RANK == -1 and torch.cuda.device_count() > 1:
+        LOGGER.warning('WARNING: DP not recommended, use torch.distributed.run for best DDP Multi-GPU results.\n'
+                       'See Multi-GPU Tutorial at https://github.com/ultralytics/yolov5/issues/475 to get started.')
+        model = torch.nn.DataParallel(model)
+
+    # SyncBatchNorm
+    if opt.sync_bn and cuda and RANK != -1:
+        model = torch.nn.SyncBatchNorm.convert_sync_batchnorm(model).to(device)
+        LOGGER.info('Using SyncBatchNorm()')
+
+    # Trainloader
+    train_loader, dataset = create_dataloader(train_path, imgsz, batch_size // WORLD_SIZE, gs, single_cls,
+                                              hyp=hyp, augment=True, cache=None if opt.cache == 'val' else opt.cache,
+                                              rect=opt.rect, rank=LOCAL_RANK, workers=workers,
+                                              image_weights=opt.image_weights, quad=opt.quad,
+                                              prefix=colorstr('train: '), shuffle=True)
+    mlc = int(np.concatenate(dataset.labels, 0)[:, 0].max())  # max label class
+    nb = len(train_loader)  # number of batches
+    assert mlc < nc, f'Label class {mlc} exceeds nc={nc} in {data}. Possible class labels are 0-{nc - 1}'
+
+    # Process 0
+    if RANK in [-1, 0]:
+        val_loader = create_dataloader(val_path, imgsz, batch_size // WORLD_SIZE * 2, gs, single_cls,
+                                       hyp=hyp, cache=None if noval else opt.cache,
+                                       rect=True, rank=-1, workers=workers * 2, pad=0.5,
+                                       prefix=colorstr('val: '))[0]
+
+        if not resume:
+            labels = np.concatenate(dataset.labels, 0)
+            # c = torch.tensor(labels[:, 0])  # classes
+            # cf = torch.bincount(c.long(), minlength=nc) + 1.  # frequency
+            # model._initialize_biases(cf.to(device))
+            if plots:
+                plot_labels(labels, names, save_dir)
+
+            # Anchors
+            if not opt.noautoanchor:
+                check_anchors(dataset, model=model, thr=hyp['anchor_t'], imgsz=imgsz)
+            model.half().float()  # pre-reduce anchor precision
+
+        callbacks.run('on_pretrain_routine_end')
+
+    # DDP mode
+    if cuda and RANK != -1:
+        model = DDP(model, device_ids=[LOCAL_RANK], output_device=LOCAL_RANK)
+
+    # Model attributes
+    nl = de_parallel(model).model[-1].nl  # number of detection layers (to scale hyps)
+    hyp['box'] *= 3 / nl  # scale to layers
+    hyp['cls'] *= nc / 80 * 3 / nl  # scale to classes and layers
+    hyp['obj'] *= (imgsz / 640) ** 2 * 3 / nl  # scale to image size and layers
+    hyp['label_smoothing'] = opt.label_smoothing
+    model.nc = nc  # attach number of classes to model
+    model.hyp = hyp  # attach hyperparameters to model
+    model.class_weights = labels_to_class_weights(dataset.labels, nc).to(device) * nc  # attach class weights
+    model.names = names
+
+    # Start training
+    t0 = time.time()
+    nw = max(round(hyp['warmup_epochs'] * nb), 1000)  # number of warmup iterations, max(3 epochs, 1k iterations)
+    # nw = min(nw, (epochs - start_epoch) / 2 * nb)  # limit warmup to < 1/2 of training
+    last_opt_step = -1
+    maps = np.zeros(nc)  # mAP per class
+    results = (0, 0, 0, 0, 0, 0, 0)  # P, R, mAP@.5, mAP@.5-.95, val_loss(box, obj, cls)
+    scheduler.last_epoch = start_epoch - 1  # do not move
+    scaler = amp.GradScaler(enabled=cuda)
+    stopper = EarlyStopping(patience=opt.patience)
+    compute_loss = ComputeLoss(model)  # init loss class
+    LOGGER.info(f'Image sizes {imgsz} train, {imgsz} val\n'
+                f'Using {train_loader.num_workers * WORLD_SIZE} dataloader workers\n'
+                f"Logging results to {colorstr('bold', save_dir)}\n"
+                f'Starting training for {epochs} epochs...')
+    for epoch in range(start_epoch, epochs):  # epoch ------------------------------------------------------------------
+        model.train()
+
+        # Update image weights (optional, single-GPU only)
+        if opt.image_weights:
+            cw = model.class_weights.cpu().numpy() * (1 - maps) ** 2 / nc  # class weights
+            iw = labels_to_image_weights(dataset.labels, nc=nc, class_weights=cw)  # image weights
+            dataset.indices = random.choices(range(dataset.n), weights=iw, k=dataset.n)  # rand weighted idx
+
+        # Update mosaic border (optional)
+        # b = int(random.uniform(0.25 * imgsz, 0.75 * imgsz + gs) // gs * gs)
+        # dataset.mosaic_border = [b - imgsz, -b]  # height, width borders
+
+        mloss = torch.zeros(3, device=device)  # mean losses
+        if RANK != -1:
+            train_loader.sampler.set_epoch(epoch)
+        pbar = enumerate(train_loader)
+        LOGGER.info(('\n' + '%10s' * 7) % ('Epoch', 'gpu_mem', 'box', 'obj', 'cls', 'labels', 'img_size'))
+        if RANK in [-1, 0]:
+            pbar = tqdm(pbar, total=nb, bar_format='{l_bar}{bar:10}{r_bar}{bar:-10b}')  # progress bar
+        optimizer.zero_grad()
+        for i, (imgs, targets, paths, _) in pbar:  # batch -------------------------------------------------------------
+            ni = i + nb * epoch  # number integrated batches (since train start)
+            imgs = imgs.to(device, non_blocking=True).float() / 255  # uint8 to float32, 0-255 to 0.0-1.0
+
+            # Warmup
+            if ni <= nw:
+                xi = [0, nw]  # x interp
+                # compute_loss.gr = np.interp(ni, xi, [0.0, 1.0])  # iou loss ratio (obj_loss = 1.0 or iou)
+                accumulate = max(1, np.interp(ni, xi, [1, nbs / batch_size]).round())
+                for j, x in enumerate(optimizer.param_groups):
+                    # bias lr falls from 0.1 to lr0, all other lrs rise from 0.0 to lr0
+                    x['lr'] = np.interp(ni, xi, [hyp['warmup_bias_lr'] if j == 2 else 0.0, x['initial_lr'] * lf(epoch)])
+                    if 'momentum' in x:
+                        x['momentum'] = np.interp(ni, xi, [hyp['warmup_momentum'], hyp['momentum']])
+
+            # Multi-scale
+            if opt.multi_scale:
+                sz = random.randrange(imgsz * 0.5, imgsz * 1.5 + gs) // gs * gs  # size
+                sf = sz / max(imgs.shape[2:])  # scale factor
+                if sf != 1:
+                    ns = [math.ceil(x * sf / gs) * gs for x in imgs.shape[2:]]  # new shape (stretched to gs-multiple)
+                    imgs = nn.functional.interpolate(imgs, size=ns, mode='bilinear', align_corners=False)
+
+            # Forward
+            with amp.autocast(enabled=cuda):
+                pred = model(imgs)  # forward
+                loss, loss_items = compute_loss(pred, targets.to(device))  # loss scaled by batch_size
+                if RANK != -1:
+                    loss *= WORLD_SIZE  # gradient averaged between devices in DDP mode
+                if opt.quad:
+                    loss *= 4.
+
+            # Backward
+            scaler.scale(loss).backward()
+
+            # Optimize
+            if ni - last_opt_step >= accumulate:
+                scaler.step(optimizer)  # optimizer.step
+                scaler.update()
+                optimizer.zero_grad()
+                if ema:
+                    ema.update(model)
+                last_opt_step = ni
+
+            # Log
+            if RANK in [-1, 0]:
+                mloss = (mloss * i + loss_items) / (i + 1)  # update mean losses
+                mem = f'{torch.cuda.memory_reserved() / 1E9 if torch.cuda.is_available() else 0:.3g}G'  # (GB)
+                pbar.set_description(('%10s' * 2 + '%10.4g' * 5) % (
+                    f'{epoch}/{epochs - 1}', mem, *mloss, targets.shape[0], imgs.shape[-1]))
+                callbacks.run('on_train_batch_end', ni, model, imgs, targets, paths, plots, opt.sync_bn)
+                if callbacks.stop_training:
+                    return
+            # end batch ------------------------------------------------------------------------------------------------
+
+        # Scheduler
+        lr = [x['lr'] for x in optimizer.param_groups]  # for loggers
+        scheduler.step()
+
+        if RANK in [-1, 0]:
+            # mAP
+            callbacks.run('on_train_epoch_end', epoch=epoch)
+            ema.update_attr(model, include=['yaml', 'nc', 'hyp', 'names', 'stride', 'class_weights'])
+            final_epoch = (epoch + 1 == epochs) or stopper.possible_stop
+            if not noval or final_epoch:  # Calculate mAP
+                results, maps, _ = val.run(data_dict,
+                                           batch_size=batch_size // WORLD_SIZE * 2,
+                                           imgsz=imgsz,
+                                           model=ema.ema,
+                                           single_cls=single_cls,
+                                           dataloader=val_loader,
+                                           save_dir=save_dir,
+                                           plots=False,
+                                           callbacks=callbacks,
+                                           compute_loss=compute_loss)
+
+            # Update best mAP
+            fi = fitness(np.array(results).reshape(1, -1))  # weighted combination of [P, R, mAP@.5, mAP@.5-.95]
+            if fi > best_fitness:
+                best_fitness = fi
+            log_vals = list(mloss) + list(results) + lr
+            callbacks.run('on_fit_epoch_end', log_vals, epoch, best_fitness, fi)
+
+            # Save model
+            if (not nosave) or (final_epoch and not evolve):  # if save
+                ckpt = {'epoch': epoch,
+                        'best_fitness': best_fitness,
+                        'model': deepcopy(de_parallel(model)).half(),
+                        'ema': deepcopy(ema.ema).half(),
+                        'updates': ema.updates,
+                        'optimizer': optimizer.state_dict(),
+                        'wandb_id': loggers.wandb.wandb_run.id if loggers.wandb else None,
+                        'date': datetime.now().isoformat()}
+
+                # Save last, best and delete
+                torch.save(ckpt, last)
+                if best_fitness == fi:
+                    torch.save(ckpt, best)
+                if (epoch > 0) and (opt.save_period > 0) and (epoch % opt.save_period == 0):
+                    torch.save(ckpt, w / f'epoch{epoch}.pt')
+                del ckpt
+                callbacks.run('on_model_save', last, epoch, final_epoch, best_fitness, fi)
+
+            # Stop Single-GPU
+            if RANK == -1 and stopper(epoch=epoch, fitness=fi):
+                break
+
+            # Stop DDP TODO: known issues shttps://github.com/ultralytics/yolov5/pull/4576
+            # stop = stopper(epoch=epoch, fitness=fi)
+            # if RANK == 0:
+            #    dist.broadcast_object_list([stop], 0)  # broadcast 'stop' to all ranks
+
+        # Stop DPP
+        # with torch_distributed_zero_first(RANK):
+        # if stop:
+        #    break  # must break all DDP ranks
+
+        # end epoch ----------------------------------------------------------------------------------------------------
+    # end training -----------------------------------------------------------------------------------------------------
+    if RANK in [-1, 0]:
+        LOGGER.info(f'\n{epoch - start_epoch + 1} epochs completed in {(time.time() - t0) / 3600:.3f} hours.')
+        for f in last, best:
+            if f.exists():
+                strip_optimizer(f)  # strip optimizers
+                if f is best:
+                    LOGGER.info(f'\nValidating {f}...')
+                    results, _, _ = val.run(data_dict,
+                                            batch_size=batch_size // WORLD_SIZE * 2,
+                                            imgsz=imgsz,
+                                            model=attempt_load(f, device).half(),
+                                            iou_thres=0.65 if is_coco else 0.60,  # best pycocotools results at 0.65
+                                            single_cls=single_cls,
+                                            dataloader=val_loader,
+                                            save_dir=save_dir,
+                                            save_json=is_coco,
+                                            verbose=True,
+                                            plots=True,
+                                            callbacks=callbacks,
+                                            compute_loss=compute_loss)  # val best model with plots
+                    if is_coco:
+                        callbacks.run('on_fit_epoch_end', list(mloss) + list(results) + lr, epoch, best_fitness, fi)
+
+        callbacks.run('on_train_end', last, best, plots, epoch, results)
+        LOGGER.info(f"Results saved to {colorstr('bold', save_dir)}")
+
+    torch.cuda.empty_cache()
+    return results
+
+
+def parse_opt(known=False):
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--weights', type=str, default=ROOT / 'VOC2007_wm/train/exp5/weights/best.pt', help='initial weights path')
+    parser.add_argument('--cfg', type=str, default='models/yolov5s_voc2007wm.yaml', help='model.yaml path')
+    parser.add_argument('--data', type=str, default=ROOT / 'data/VOC.yaml', help='dataset.yaml path')
+    parser.add_argument('--hyp', type=str, default=ROOT / 'data/hyps/hyp.scratch-low.yaml', help='hyperparameters path')
+    parser.add_argument('--epochs', type=int, default=10)
+    parser.add_argument('--batch-size', type=int, default=16, help='total batch size for all GPUs, -1 for autobatch')
+    parser.add_argument('--imgsz', '--img', '--img-size', type=int, default=640, help='train, val image size (pixels)')
+    parser.add_argument('--rect', action='store_true', help='rectangular training')
+    parser.add_argument('--resume', nargs='?', const=True, default=False, help='resume most recent training')
+    parser.add_argument('--nosave', action='store_true', help='only save final checkpoint')
+    parser.add_argument('--noval', action='store_true', help='only validate final epoch')
+    parser.add_argument('--noautoanchor', action='store_true', help='disable AutoAnchor')
+    parser.add_argument('--evolve', type=int, nargs='?', const=300, help='evolve hyperparameters for x generations')
+    parser.add_argument('--bucket', type=str, default='', help='gsutil bucket')
+    parser.add_argument('--cache', type=str, nargs='?', const='ram', help='--cache images in "ram" (default) or "disk"')
+    parser.add_argument('--image-weights', action='store_true', help='use weighted image selection for training')
+    parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')
+    parser.add_argument('--multi-scale', action='store_true', help='vary img-size +/- 50%%')
+    parser.add_argument('--single-cls', action='store_true', help='train multi-class data as single-class')
+    parser.add_argument('--optimizer', type=str, choices=['SGD', 'Adam', 'AdamW'], default='SGD', help='optimizer')
+    parser.add_argument('--sync-bn', action='store_true', help='use SyncBatchNorm, only available in DDP mode')
+    parser.add_argument('--workers', type=int, default=8, help='max dataloader workers (per RANK in DDP mode)')
+    parser.add_argument('--project', default=ROOT / 'VOC2007_wm/train', help='save to project/name')
+    parser.add_argument('--name', default='exp', help='save to project/name')
+    parser.add_argument('--exist-ok', action='store_true', help='existing project/name ok, do not increment')
+    parser.add_argument('--quad', action='store_true', help='quad dataloader')
+    parser.add_argument('--cos-lr', action='store_true', help='cosine LR scheduler')
+    parser.add_argument('--label-smoothing', type=float, default=0.0, help='Label smoothing epsilon')
+    parser.add_argument('--patience', type=int, default=100, help='EarlyStopping patience (epochs without improvement)')
+    parser.add_argument('--freeze', nargs='+', type=int, default=[0], help='Freeze layers: backbone=10, first3=0 1 2')
+    parser.add_argument('--save-period', type=int, default=-1, help='Save checkpoint every x epochs (disabled if < 1)')
+    parser.add_argument('--local_rank', type=int, default=-1, help='DDP parameter, do not modify')
+
+    # Weights & Biases arguments
+    parser.add_argument('--entity', default=None, help='W&B: Entity')
+    parser.add_argument('--upload_dataset', nargs='?', const=True, default=False, help='W&B: Upload data, "val" option')
+    parser.add_argument('--bbox_interval', type=int, default=-1, help='W&B: Set bounding-box image logging interval')
+    parser.add_argument('--artifact_alias', type=str, default='latest', help='W&B: Version of dataset artifact to use')
+
+    opt = parser.parse_known_args()[0] if known else parser.parse_args()
+    return opt
+
+
+def main(opt, callbacks=Callbacks()):
+    # Checks
+    if RANK in [-1, 0]:
+        print_args(FILE.stem, opt)
+        check_git_status()
+        check_requirements(exclude=['thop'])
+
+    # Resume
+    if opt.resume and not check_wandb_resume(opt) and not opt.evolve:  # resume an interrupted run
+        ckpt = opt.resume if isinstance(opt.resume, str) else get_latest_run()  # specified or most recent path
+        assert os.path.isfile(ckpt), 'ERROR: --resume checkpoint does not exist'
+        with open(Path(ckpt).parent.parent / 'opt.yaml', errors='ignore') as f:
+            opt = argparse.Namespace(**yaml.safe_load(f))  # replace
+        opt.cfg, opt.weights, opt.resume = '', ckpt, True  # reinstate
+        LOGGER.info(f'Resuming training from {ckpt}')
+    else:
+        opt.data, opt.cfg, opt.hyp, opt.weights, opt.project = \
+            check_file(opt.data), check_yaml(opt.cfg), check_yaml(opt.hyp), str(opt.weights), str(opt.project)  # checks
+        assert len(opt.cfg) or len(opt.weights), 'either --cfg or --weights must be specified'
+        if opt.evolve:
+            if opt.project == str(ROOT / 'runs/train'):  # if default project name, rename to runs/evolve
+                opt.project = str(ROOT / 'runs/evolve')
+            opt.exist_ok, opt.resume = opt.resume, False  # pass resume to exist_ok and disable resume
+        opt.save_dir = str(increment_path(Path(opt.project) / opt.name, exist_ok=opt.exist_ok))
+
+    # DDP mode
+    device = select_device(opt.device, batch_size=opt.batch_size)
+    if LOCAL_RANK != -1:
+        msg = 'is not compatible with YOLOv5 Multi-GPU DDP training'
+        assert not opt.image_weights, f'--image-weights {msg}'
+        assert not opt.evolve, f'--evolve {msg}'
+        assert opt.batch_size != -1, f'AutoBatch with --batch-size -1 {msg}, please pass a valid --batch-size'
+        assert opt.batch_size % WORLD_SIZE == 0, f'--batch-size {opt.batch_size} must be multiple of WORLD_SIZE'
+        assert torch.cuda.device_count() > LOCAL_RANK, 'insufficient CUDA devices for DDP command'
+        torch.cuda.set_device(LOCAL_RANK)
+        device = torch.device('cuda', LOCAL_RANK)
+        dist.init_process_group(backend="nccl" if dist.is_nccl_available() else "gloo")
+
+    # Train
+    if not opt.evolve:
+        train(opt.hyp, opt, device, callbacks)
+        if WORLD_SIZE > 1 and RANK == 0:
+            LOGGER.info('Destroying process group... ')
+            dist.destroy_process_group()
+
+    # Evolve hyperparameters (optional)
+    else:
+        # Hyperparameter evolution metadata (mutation scale 0-1, lower_limit, upper_limit)
+        meta = {'lr0': (1, 1e-5, 1e-1),  # initial learning rate (SGD=1E-2, Adam=1E-3)
+                'lrf': (1, 0.01, 1.0),  # final OneCycleLR learning rate (lr0 * lrf)
+                'momentum': (0.3, 0.6, 0.98),  # SGD momentum/Adam beta1
+                'weight_decay': (1, 0.0, 0.001),  # optimizer weight decay
+                'warmup_epochs': (1, 0.0, 5.0),  # warmup epochs (fractions ok)
+                'warmup_momentum': (1, 0.0, 0.95),  # warmup initial momentum
+                'warmup_bias_lr': (1, 0.0, 0.2),  # warmup initial bias lr
+                'box': (1, 0.02, 0.2),  # box loss gain
+                'cls': (1, 0.2, 4.0),  # cls loss gain
+                'cls_pw': (1, 0.5, 2.0),  # cls BCELoss positive_weight
+                'obj': (1, 0.2, 4.0),  # obj loss gain (scale with pixels)
+                'obj_pw': (1, 0.5, 2.0),  # obj BCELoss positive_weight
+                'iou_t': (0, 0.1, 0.7),  # IoU training threshold
+                'anchor_t': (1, 2.0, 8.0),  # anchor-multiple threshold
+                'anchors': (2, 2.0, 10.0),  # anchors per output grid (0 to ignore)
+                'fl_gamma': (0, 0.0, 2.0),  # focal loss gamma (efficientDet default gamma=1.5)
+                'hsv_h': (1, 0.0, 0.1),  # image HSV-Hue augmentation (fraction)
+                'hsv_s': (1, 0.0, 0.9),  # image HSV-Saturation augmentation (fraction)
+                'hsv_v': (1, 0.0, 0.9),  # image HSV-Value augmentation (fraction)
+                'degrees': (1, 0.0, 45.0),  # image rotation (+/- deg)
+                'translate': (1, 0.0, 0.9),  # image translation (+/- fraction)
+                'scale': (1, 0.0, 0.9),  # image scale (+/- gain)
+                'shear': (1, 0.0, 10.0),  # image shear (+/- deg)
+                'perspective': (0, 0.0, 0.001),  # image perspective (+/- fraction), range 0-0.001
+                'flipud': (1, 0.0, 1.0),  # image flip up-down (probability)
+                'fliplr': (0, 0.0, 1.0),  # image flip left-right (probability)
+                'mosaic': (1, 0.0, 1.0),  # image mixup (probability)
+                'mixup': (1, 0.0, 1.0),  # image mixup (probability)
+                'copy_paste': (1, 0.0, 1.0)}  # segment copy-paste (probability)
+
+        with open(opt.hyp, errors='ignore') as f:
+            hyp = yaml.safe_load(f)  # load hyps dict
+            if 'anchors' not in hyp:  # anchors commented in hyp.yaml
+                hyp['anchors'] = 3
+        opt.noval, opt.nosave, save_dir = True, True, Path(opt.save_dir)  # only val/save final epoch
+        # ei = [isinstance(x, (int, float)) for x in hyp.values()]  # evolvable indices
+        evolve_yaml, evolve_csv = save_dir / 'hyp_evolve.yaml', save_dir / 'evolve.csv'
+        if opt.bucket:
+            os.system(f'gsutil cp gs://{opt.bucket}/evolve.csv {evolve_csv}')  # download evolve.csv if exists
+
+        for _ in range(opt.evolve):  # generations to evolve
+            if evolve_csv.exists():  # if evolve.csv exists: select best hyps and mutate
+                # Select parent(s)
+                parent = 'single'  # parent selection method: 'single' or 'weighted'
+                x = np.loadtxt(evolve_csv, ndmin=2, delimiter=',', skiprows=1)
+                n = min(5, len(x))  # number of previous results to consider
+                x = x[np.argsort(-fitness(x))][:n]  # top n mutations
+                w = fitness(x) - fitness(x).min() + 1E-6  # weights (sum > 0)
+                if parent == 'single' or len(x) == 1:
+                    # x = x[random.randint(0, n - 1)]  # random selection
+                    x = x[random.choices(range(n), weights=w)[0]]  # weighted selection
+                elif parent == 'weighted':
+                    x = (x * w.reshape(n, 1)).sum(0) / w.sum()  # weighted combination
+
+                # Mutate
+                mp, s = 0.8, 0.2  # mutation probability, sigma
+                npr = np.random
+                npr.seed(int(time.time()))
+                g = np.array([meta[k][0] for k in hyp.keys()])  # gains 0-1
+                ng = len(meta)
+                v = np.ones(ng)
+                while all(v == 1):  # mutate until a change occurs (prevent duplicates)
+                    v = (g * (npr.random(ng) < mp) * npr.randn(ng) * npr.random() * s + 1).clip(0.3, 3.0)
+                for i, k in enumerate(hyp.keys()):  # plt.hist(v.ravel(), 300)
+                    hyp[k] = float(x[i + 7] * v[i])  # mutate
+
+            # Constrain to limits
+            for k, v in meta.items():
+                hyp[k] = max(hyp[k], v[1])  # lower limit
+                hyp[k] = min(hyp[k], v[2])  # upper limit
+                hyp[k] = round(hyp[k], 5)  # significant digits
+
+            # Train mutation
+            results = train(hyp.copy(), opt, device, callbacks)
+            callbacks = Callbacks()
+            # Write mutation results
+            print_mutation(results, hyp.copy(), save_dir, opt.bucket)
+
+        # Plot results
+        plot_evolve(evolve_csv)
+        LOGGER.info(f'Hyperparameter evolution finished {opt.evolve} generations\n'
+                    f"Results saved to {colorstr('bold', save_dir)}\n"
+                    f'Usage example: $ python train.py --hyp {evolve_yaml}')
+
+
+def run(**kwargs):
+    # Usage: import train; train.run(data='coco128.yaml', imgsz=320, weights='yolov5m.pt')
+    opt = parse_opt(True)
+    for k, v in kwargs.items():
+        setattr(opt, k, v)
+    main(opt)
+    return opt
+
+
+if __name__ == "__main__":
+    opt = parse_opt()
+    main(opt)

train_sparsity.py

@@ -0,0 +1,680 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+"""
+Train a YOLOv5 model on a custom dataset.
+
+Models and datasets download automatically from the latest YOLOv5 release.
+Models: https://github.com/ultralytics/yolov5/tree/master/models
+Datasets: https://github.com/ultralytics/yolov5/tree/master/data
+Tutorial: https://github.com/ultralytics/yolov5/wiki/Train-Custom-Data
+
+Usage:
+    $ python path/to/train.py --data coco128.yaml --weights yolov5s.pt --img 640  # from pretrained (RECOMMENDED)
+    $ python path/to/train.py --data coco128.yaml --weights '' --cfg yolov5s.yaml --img 640  # from scratch
+"""
+
+import argparse
+import math
+import os
+import random
+import sys
+import time
+from copy import deepcopy
+from datetime import datetime
+from pathlib import Path
+
+import numpy as np
+import torch
+import torch.distributed as dist
+import torch.nn as nn
+import yaml
+from torch.cuda import amp
+from torch.nn.parallel import DistributedDataParallel as DDP
+from torch.optim import SGD, Adam, AdamW, lr_scheduler
+from tqdm import tqdm
+
+FILE = Path(__file__).resolve()
+ROOT = FILE.parents[0]  # YOLOv5 root directory
+if str(ROOT) not in sys.path:
+    sys.path.append(str(ROOT))  # add ROOT to PATH
+ROOT = Path(os.path.relpath(ROOT, Path.cwd()))  # relative
+
+import val  # for end-of-epoch mAP
+from models.experimental import attempt_load
+from models.yolo import Model
+from utils.autoanchor import check_anchors
+from utils.autobatch import check_train_batch_size
+from utils.callbacks import Callbacks
+from utils.datasets import create_dataloader
+from utils.downloads import attempt_download
+from utils.general import (LOGGER, check_dataset, check_file, check_git_status, check_img_size, check_requirements,
+                           check_suffix, check_yaml, colorstr, get_latest_run, increment_path, init_seeds,
+                           intersect_dicts, labels_to_class_weights, labels_to_image_weights, methods, one_cycle,
+                           print_args, print_mutation, strip_optimizer)
+from utils.loggers import Loggers
+from utils.loggers.wandb.wandb_utils import check_wandb_resume
+from utils.loss import ComputeLoss
+from utils.metrics import fitness
+from utils.plots import plot_evolve, plot_labels
+from utils.torch_utils import EarlyStopping, ModelEMA, de_parallel, select_device, torch_distributed_zero_first
+from models.common import Bottleneck
+
+LOCAL_RANK = int(os.getenv('LOCAL_RANK', -1))  # https://pytorch.org/docs/stable/elastic/run.html
+RANK = int(os.getenv('RANK', -1))
+WORLD_SIZE = int(os.getenv('WORLD_SIZE', 1))
+
+
+def train(hyp,  # path/to/hyp.yaml or hyp dictionary
+          opt,
+          device,
+          callbacks
+          ):
+    save_dir, epochs, batch_size, weights, single_cls, evolve, data, cfg, resume, noval, nosave, workers, freeze = \
+        Path(opt.save_dir), opt.epochs, opt.batch_size, opt.weights, opt.single_cls, opt.evolve, opt.data, opt.cfg, \
+        opt.resume, opt.noval, opt.nosave, opt.workers, opt.freeze
+
+    # Directories
+    w = save_dir / 'weights'  # weights dir
+    (w.parent if evolve else w).mkdir(parents=True, exist_ok=True)  # make dir
+    last, best = w / 'last.pt', w / 'best.pt'
+
+    # Hyperparameters
+    if isinstance(hyp, str):
+        with open(hyp, errors='ignore') as f:
+            hyp = yaml.safe_load(f)  # load hyps dict
+    LOGGER.info(colorstr('hyperparameters: ') + ', '.join(f'{k}={v}' for k, v in hyp.items()))
+
+    # Save run settings
+    if not evolve:
+        with open(save_dir / 'hyp.yaml', 'w') as f:
+            yaml.safe_dump(hyp, f, sort_keys=False)
+        with open(save_dir / 'opt.yaml', 'w') as f:
+            yaml.safe_dump(vars(opt), f, sort_keys=False)
+
+    # Loggers
+    data_dict = None
+    if RANK in [-1, 0]:
+        loggers = Loggers(save_dir, weights, opt, hyp, LOGGER)  # loggers instance
+        if loggers.wandb:
+            data_dict = loggers.wandb.data_dict
+            if resume:
+                weights, epochs, hyp, batch_size = opt.weights, opt.epochs, opt.hyp, opt.batch_size
+
+        # Register actions
+        for k in methods(loggers):
+            callbacks.register_action(k, callback=getattr(loggers, k))
+
+    # Config
+    plots = not evolve  # create plots
+    cuda = device.type != 'cpu'
+    init_seeds(1 + RANK)
+    with torch_distributed_zero_first(LOCAL_RANK):
+        data_dict = data_dict or check_dataset(data)  # check if None
+    train_path, val_path = data_dict['train'], data_dict['val']
+    nc = 1 if single_cls else int(data_dict['nc'])  # number of classes
+    names = ['item'] if single_cls and len(data_dict['names']) != 1 else data_dict['names']  # class names
+    assert len(names) == nc, f'{len(names)} names found for nc={nc} dataset in {data}'  # check
+    is_coco = isinstance(val_path, str) and val_path.endswith('coco/val2017.txt')  # COCO dataset
+
+    # Model
+    check_suffix(weights, '.pt')  # check weights
+    pretrained = weights.endswith('.pt')
+    if pretrained:
+        with torch_distributed_zero_first(LOCAL_RANK):
+            weights = attempt_download(weights)  # download if not found locally
+        ckpt = torch.load(weights, map_location='cpu')  # load checkpoint to CPU to avoid CUDA memory leak
+        model = Model(cfg or ckpt['model'].yaml, ch=3, nc=nc, anchors=hyp.get('anchors')).to(device)  # create
+        exclude = ['anchor'] if (cfg or hyp.get('anchors')) and not resume else []  # exclude keys
+        csd = ckpt['model'].float().state_dict()  # checkpoint state_dict as FP32
+        csd = intersect_dicts(csd, model.state_dict(), exclude=exclude)  # intersect
+        model.load_state_dict(csd, strict=False)  # load
+        LOGGER.info(f'Transferred {len(csd)}/{len(model.state_dict())} items from {weights}')  # report
+    else:
+        model = Model(cfg, ch=3, nc=nc, anchors=hyp.get('anchors')).to(device)  # create
+
+    # Freeze
+    freeze = [f'model.{x}.' for x in (freeze if len(freeze) > 1 else range(freeze[0]))]  # layers to freeze
+    for k, v in model.named_parameters():
+        v.requires_grad = True  # train all layers
+        if any(x in k for x in freeze):
+            LOGGER.info(f'freezing {k}')
+            v.requires_grad = False
+
+    # Image size
+    gs = max(int(model.stride.max()), 32)  # grid size (max stride)
+    imgsz = check_img_size(opt.imgsz, gs, floor=gs * 2)  # verify imgsz is gs-multiple
+
+    # Batch size
+    if RANK == -1 and batch_size == -1:  # single-GPU only, estimate best batch size
+        batch_size = check_train_batch_size(model, imgsz)
+        loggers.on_params_update({"batch_size": batch_size})
+
+    # Optimizer
+    nbs = 64  # nominal batch size
+    accumulate = max(round(nbs / batch_size), 1)  # accumulate loss before optimizing
+    hyp['weight_decay'] *= batch_size * accumulate / nbs  # scale weight_decay
+    LOGGER.info(f"Scaled weight_decay = {hyp['weight_decay']}")
+
+    g0, g1, g2 = [], [], []  # optimizer parameter groups
+    for v in model.modules():
+        if hasattr(v, 'bias') and isinstance(v.bias, nn.Parameter):  # bias
+            g2.append(v.bias)
+        if isinstance(v, nn.BatchNorm2d):  # weight (no decay)
+            g0.append(v.weight)
+        elif hasattr(v, 'weight') and isinstance(v.weight, nn.Parameter):  # weight (with decay)
+            g1.append(v.weight)
+
+    if opt.optimizer == 'Adam':
+        optimizer = Adam(g0, lr=hyp['lr0'], betas=(hyp['momentum'], 0.999))  # adjust beta1 to momentum
+    elif opt.optimizer == 'AdamW':
+        optimizer = AdamW(g0, lr=hyp['lr0'], betas=(hyp['momentum'], 0.999))  # adjust beta1 to momentum
+    else:
+        optimizer = SGD(g0, lr=hyp['lr0'], momentum=hyp['momentum'], nesterov=True)
+
+    optimizer.add_param_group({'params': g1, 'weight_decay': hyp['weight_decay']})  # add g1 with weight_decay
+    optimizer.add_param_group({'params': g2})  # add g2 (biases)
+    LOGGER.info(f"{colorstr('optimizer:')} {type(optimizer).__name__} with parameter groups "
+                f"{len(g0)} weight (no decay), {len(g1)} weight, {len(g2)} bias")
+    del g0, g1, g2
+
+    # Scheduler
+    if opt.cos_lr:
+        lf = one_cycle(1, hyp['lrf'], epochs)  # cosine 1->hyp['lrf']
+    else:
+        lf = lambda x: (1 - x / epochs) * (1.0 - hyp['lrf']) + hyp['lrf']  # linear
+    scheduler = lr_scheduler.LambdaLR(optimizer, lr_lambda=lf)  # plot_lr_scheduler(optimizer, scheduler, epochs)
+
+    # EMA
+    ema = ModelEMA(model) if RANK in [-1, 0] else None
+
+    # Resume
+    start_epoch, best_fitness = 0, 0.0
+    if pretrained:
+        # Optimizer
+        if ckpt['optimizer'] is not None:
+            optimizer.load_state_dict(ckpt['optimizer'])
+            best_fitness = ckpt['best_fitness']
+
+        # EMA
+        if ema and ckpt.get('ema'):
+            ema.ema.load_state_dict(ckpt['ema'].float().state_dict())
+            ema.updates = ckpt['updates']
+
+        # Epochs
+        start_epoch = ckpt['epoch'] + 1
+        if resume:
+            assert start_epoch > 0, f'{weights} training to {epochs} epochs is finished, nothing to resume.'
+        if epochs < start_epoch:
+            LOGGER.info(f"{weights} has been trained for {ckpt['epoch']} epochs. Fine-tuning for {epochs} more epochs.")
+            epochs += ckpt['epoch']  # finetune additional epochs
+
+        del ckpt, csd
+
+    # DP mode
+    if cuda and RANK == -1 and torch.cuda.device_count() > 1:
+        LOGGER.warning('WARNING: DP not recommended, use torch.distributed.run for best DDP Multi-GPU results.\n'
+                       'See Multi-GPU Tutorial at https://github.com/ultralytics/yolov5/issues/475 to get started.')
+        model = torch.nn.DataParallel(model)
+
+    # SyncBatchNorm
+    if opt.sync_bn and cuda and RANK != -1:
+        model = torch.nn.SyncBatchNorm.convert_sync_batchnorm(model).to(device)
+        LOGGER.info('Using SyncBatchNorm()')
+    print(model)
+    # Trainloader
+    train_loader, dataset = create_dataloader(train_path, imgsz, batch_size // WORLD_SIZE, gs, single_cls,
+                                              hyp=hyp, augment=True, cache=None if opt.cache == 'val' else opt.cache,
+                                              rect=opt.rect, rank=LOCAL_RANK, workers=workers,
+                                              image_weights=opt.image_weights, quad=opt.quad,
+                                              prefix=colorstr('train: '), shuffle=True)
+    mlc = int(np.concatenate(dataset.labels, 0)[:, 0].max())  # max label class
+    nb = len(train_loader)  # number of batches
+    assert mlc < nc, f'Label class {mlc} exceeds nc={nc} in {data}. Possible class labels are 0-{nc - 1}'
+
+    # Process 0
+    if RANK in [-1, 0]:
+        val_loader = create_dataloader(val_path, imgsz, batch_size // WORLD_SIZE * 2, gs, single_cls,
+                                       hyp=hyp, cache=None if noval else opt.cache,
+                                       rect=True, rank=-1, workers=workers * 2, pad=0.5,
+                                       prefix=colorstr('val: '))[0]
+
+        if not resume:
+            labels = np.concatenate(dataset.labels, 0)
+            # c = torch.tensor(labels[:, 0])  # classes
+            # cf = torch.bincount(c.long(), minlength=nc) + 1.  # frequency
+            # model._initialize_biases(cf.to(device))
+            if plots:
+                plot_labels(labels, names, save_dir)
+
+            # Anchors
+            if not opt.noautoanchor:
+                check_anchors(dataset, model=model, thr=hyp['anchor_t'], imgsz=imgsz)
+            model.half().float()  # pre-reduce anchor precision
+
+        callbacks.run('on_pretrain_routine_end')
+
+    # DDP mode
+    if cuda and RANK != -1:
+        model = DDP(model, device_ids=[LOCAL_RANK], output_device=LOCAL_RANK)
+
+    # Model attributes
+    nl = de_parallel(model).model[-1].nl  # number of detection layers (to scale hyps)
+    hyp['box'] *= 3 / nl  # scale to layers
+    hyp['cls'] *= nc / 80 * 3 / nl  # scale to classes and layers
+    hyp['obj'] *= (imgsz / 640) ** 2 * 3 / nl  # scale to image size and layers
+    hyp['label_smoothing'] = opt.label_smoothing
+    model.nc = nc  # attach number of classes to model
+    model.hyp = hyp  # attach hyperparameters to model
+    model.class_weights = labels_to_class_weights(dataset.labels, nc).to(device) * nc  # attach class weights
+    model.names = names
+
+    # Start training
+    t0 = time.time()
+    nw = max(round(hyp['warmup_epochs'] * nb), 1000)  # number of warmup iterations, max(3 epochs, 1k iterations)
+    # nw = min(nw, (epochs - start_epoch) / 2 * nb)  # limit warmup to < 1/2 of training
+    last_opt_step = -1
+    maps = np.zeros(nc)  # mAP per class
+    results = (0, 0, 0, 0, 0, 0, 0)  # P, R, mAP@.5, mAP@.5-.95, val_loss(box, obj, cls)
+    scheduler.last_epoch = start_epoch - 1  # do not move
+    scaler = amp.GradScaler(enabled=cuda)
+    stopper = EarlyStopping(patience=opt.patience)
+    compute_loss = ComputeLoss(model)  # init loss class
+    LOGGER.info(f'Image sizes {imgsz} train, {imgsz} val\n'
+                f'Using {train_loader.num_workers * WORLD_SIZE} dataloader workers\n'
+                f"Logging results to {colorstr('bold', save_dir)}\n"
+                f'Starting training for {epochs} epochs...')
+    for epoch in range(start_epoch, epochs):  # epoch ------------------------------------------------------------------
+        model.train()
+
+        # Update image weights (optional, single-GPU only)
+        if opt.image_weights:
+            cw = model.class_weights.cpu().numpy() * (1 - maps) ** 2 / nc  # class weights
+            iw = labels_to_image_weights(dataset.labels, nc=nc, class_weights=cw)  # image weights
+            dataset.indices = random.choices(range(dataset.n), weights=iw, k=dataset.n)  # rand weighted idx
+
+        # Update mosaic border (optional)
+        # b = int(random.uniform(0.25 * imgsz, 0.75 * imgsz + gs) // gs * gs)
+        # dataset.mosaic_border = [b - imgsz, -b]  # height, width borders
+
+        mloss = torch.zeros(3, device=device)  # mean losses
+        if RANK != -1:
+            train_loader.sampler.set_epoch(epoch)
+        pbar = enumerate(train_loader)
+        LOGGER.info(('\n' + '%10s' * 7) % ('Epoch', 'gpu_mem', 'box', 'obj', 'cls', 'labels', 'img_size'))
+        if RANK in [-1, 0]:
+            pbar = tqdm(pbar, total=nb, bar_format='{l_bar}{bar:10}{r_bar}{bar:-10b}')  # progress bar
+        optimizer.zero_grad()
+        for i, (imgs, targets, paths, _) in pbar:  # batch -------------------------------------------------------------
+            ni = i + nb * epoch  # number integrated batches (since train start)
+            imgs = imgs.to(device, non_blocking=True).float() / 255  # uint8 to float32, 0-255 to 0.0-1.0
+
+            # Warmup
+            if ni <= nw:
+                xi = [0, nw]  # x interp
+                # compute_loss.gr = np.interp(ni, xi, [0.0, 1.0])  # iou loss ratio (obj_loss = 1.0 or iou)
+                accumulate = max(1, np.interp(ni, xi, [1, nbs / batch_size]).round())
+                for j, x in enumerate(optimizer.param_groups):
+                    # bias lr falls from 0.1 to lr0, all other lrs rise from 0.0 to lr0
+                    x['lr'] = np.interp(ni, xi, [hyp['warmup_bias_lr'] if j == 2 else 0.0, x['initial_lr'] * lf(epoch)])
+                    if 'momentum' in x:
+                        x['momentum'] = np.interp(ni, xi, [hyp['warmup_momentum'], hyp['momentum']])
+
+            # Multi-scale
+            if opt.multi_scale:
+                sz = random.randrange(imgsz * 0.5, imgsz * 1.5 + gs) // gs * gs  # size
+                sf = sz / max(imgs.shape[2:])  # scale factor
+                if sf != 1:
+                    ns = [math.ceil(x * sf / gs) * gs for x in imgs.shape[2:]]  # new shape (stretched to gs-multiple)
+                    imgs = nn.functional.interpolate(imgs, size=ns, mode='bilinear', align_corners=False)
+
+            # Forward
+            with amp.autocast(enabled=cuda):
+	            pred = model(imgs)  # forward
+	            loss, loss_items = compute_loss(pred, targets.to(device))  # loss scaled by batch_size
+	            if RANK != -1:
+	                loss *= WORLD_SIZE  # gradient averaged between devices in DDP mode
+	            if opt.quad:
+	                loss *= 4.
+
+            # Backward
+            # scaler.scale(loss).backward()
+            loss.backward()
+            # # ============================= sparsity training ========================== #
+            srtmp = opt.sr*(1 - 0.9*epoch/epochs)
+            if opt.st:
+                ignore_bn_list = []
+                for k, m in model.named_modules():
+                    if isinstance(m, Bottleneck):
+                        if m.add:
+                            ignore_bn_list.append(k.rsplit(".", 2)[0] + ".cv1.bn")
+                            ignore_bn_list.append(k + '.cv1.bn')
+                            ignore_bn_list.append(k + '.cv2.bn')
+                    if isinstance(m, nn.BatchNorm2d) and (k not in ignore_bn_list):
+                        m.weight.grad.data.add_(srtmp * torch.sign(m.weight.data))  # L1
+                        m.bias.grad.data.add_(opt.sr*10 * torch.sign(m.bias.data))  # L1
+            # # ============================= sparsity training ========================== #
+
+            # Optimize
+            # if ni - last_opt_step >= accumulate:
+            optimizer.step()
+            # scaler.step(optimizer)  # optimizer.step
+            # scaler.update()
+            optimizer.zero_grad()
+            if ema:
+                ema.update(model)
+            # last_opt_step = ni
+
+            # Log
+            if RANK in [-1, 0]:
+                mloss = (mloss * i + loss_items) / (i + 1)  # update mean losses
+                mem = f'{torch.cuda.memory_reserved() / 1E9 if torch.cuda.is_available() else 0:.3g}G'  # (GB)
+                pbar.set_description(('%10s' * 2 + '%10.4g' * 5) % (
+                    f'{epoch}/{epochs - 1}', mem, *mloss, targets.shape[0], imgs.shape[-1]))
+                callbacks.run('on_train_batch_end', ni, model, imgs, targets, paths, plots, opt.sync_bn)
+                if callbacks.stop_training:
+                    return
+            # end batch ------------------------------------------------------------------------------------------------
+
+        # Scheduler
+        lr = [x['lr'] for x in optimizer.param_groups]  # for loggers
+        scheduler.step()
+
+        # =============== show bn weights ===================== #
+        module_list = []
+        for i, layer in model.named_modules():
+            if isinstance(layer, nn.BatchNorm2d) and i not in ignore_bn_list:
+                bnw = layer.state_dict()['weight']
+                bnb = layer.state_dict()['bias']
+                module_list.append(bnw)
+        size_list = [idx.data.shape[0] for idx in module_list]
+
+        bn_weights = torch.zeros(sum(size_list))
+        bnb_weights = torch.zeros(sum(size_list))
+        index = 0
+        for idx, size in enumerate(size_list):
+            bn_weights[index:(index + size)] = module_list[idx].data.abs().clone()            
+            index += size
+
+        if RANK in [-1, 0]:
+            # mAP
+            callbacks.run('on_train_epoch_end', epoch=epoch)
+            ema.update_attr(model, include=['yaml', 'nc', 'hyp', 'names', 'stride', 'class_weights'])
+            final_epoch = (epoch + 1 == epochs) or stopper.possible_stop
+            if not noval or final_epoch:  # Calculate mAP
+                results, maps, _ = val.run(data_dict,
+                                           batch_size=batch_size // WORLD_SIZE * 2,
+                                           imgsz=imgsz,
+                                           model=ema.ema,
+                                           single_cls=single_cls,
+                                           dataloader=val_loader,
+                                           save_dir=save_dir,
+                                           plots=False,
+                                           callbacks=callbacks,
+                                           compute_loss=compute_loss)
+
+            # Update best mAP
+            fi = fitness(np.array(results).reshape(1, -1))  # weighted combination of [P, R, mAP@.5, mAP@.5-.95]
+            if fi > best_fitness:
+                best_fitness = fi
+            #log_vals = list(mloss) + list(results) + lr + [srtmp]
+            log_vals = list(mloss) + list(results) + lr
+            callbacks.run('on_fit_epoch_end', log_vals, epoch, best_fitness, fi)
+            callbacks.run('on_fit_epoch_end_prune', bn_weights.numpy(), epoch)
+
+            # Save model
+            if (not nosave) or (final_epoch and not evolve):  # if save
+                ckpt = {'epoch': epoch,
+                        'best_fitness': best_fitness,
+                        'model': deepcopy(de_parallel(model)).half(),
+                        'ema': deepcopy(ema.ema).half(),
+                        'updates': ema.updates,
+                        'optimizer': optimizer.state_dict(),
+                        'wandb_id': loggers.wandb.wandb_run.id if loggers.wandb else None,
+                        'date': datetime.now().isoformat()}
+
+                # Save last, best and delete
+                torch.save(ckpt, last)
+                if best_fitness == fi:
+                    torch.save(ckpt, best)
+                if (epoch > 0) and (opt.save_period > 0) and (epoch % opt.save_period == 0):
+                    torch.save(ckpt, w / f'epoch{epoch}.pt')
+                del ckpt
+                callbacks.run('on_model_save', last, epoch, final_epoch, best_fitness, fi)
+
+            # Stop Single-GPU
+            if RANK == -1 and stopper(epoch=epoch, fitness=fi):
+                break
+
+            # Stop DDP TODO: known issues shttps://github.com/ultralytics/yolov5/pull/4576
+            # stop = stopper(epoch=epoch, fitness=fi)
+            # if RANK == 0:
+            #    dist.broadcast_object_list([stop], 0)  # broadcast 'stop' to all ranks
+
+        # Stop DPP
+        # with torch_distributed_zero_first(RANK):
+        # if stop:
+        #    break  # must break all DDP ranks
+
+        # end epoch ----------------------------------------------------------------------------------------------------
+    # end training -----------------------------------------------------------------------------------------------------
+    if RANK in [-1, 0]:
+        LOGGER.info(f'\n{epoch - start_epoch + 1} epochs completed in {(time.time() - t0) / 3600:.3f} hours.')
+        for f in last, best:
+            if f.exists():
+                strip_optimizer(f)  # strip optimizers
+                if f is best:
+                    LOGGER.info(f'\nValidating {f}...')
+                    results, _, _ = val.run(data_dict,
+                                            batch_size=batch_size // WORLD_SIZE * 2,
+                                            imgsz=imgsz,
+                                            model=attempt_load(f, device).half(),
+                                            iou_thres=0.65 if is_coco else 0.60,  # best pycocotools results at 0.65
+                                            single_cls=single_cls,
+                                            dataloader=val_loader,
+                                            save_dir=save_dir,
+                                            save_json=is_coco,
+                                            verbose=True,
+                                            plots=True,
+                                            callbacks=callbacks,
+                                            compute_loss=compute_loss)  # val best model with plots
+                    if is_coco:
+                        callbacks.run('on_fit_epoch_end', list(mloss) + list(results) + lr, epoch, best_fitness, fi)
+
+        callbacks.run('on_train_end', last, best, plots, epoch, results)
+        LOGGER.info(f"Results saved to {colorstr('bold', save_dir)}")
+
+    torch.cuda.empty_cache()
+    return results
+
+
+def parse_opt(known=False):
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--st', action='store_true',default=True, help='train with L1 sparsity normalization')
+    parser.add_argument('--sr', type=float, default=0.0001, help='L1 normal sparse rate')
+    parser.add_argument('--weights', type=str, default=ROOT / 'VOC2007_wm/train/exp5/weights/last.pt', help='initial weights path')
+    parser.add_argument('--cfg', type=str, default='models/yolov5s.yaml', help='model.yaml path')
+    parser.add_argument('--data', type=str, default=ROOT / 'data/VOC.yaml', help='dataset.yaml path')
+    parser.add_argument('--hyp', type=str, default=ROOT / 'data/hyps/hyp.scratch-low.yaml', help='hyperparameters path')
+    parser.add_argument('--epochs', type=int, default=300)
+    parser.add_argument('--batch-size', type=int, default=16, help='total batch size for all GPUs, -1 for autobatch')
+    parser.add_argument('--imgsz', '--img', '--img-size', type=int, default=640, help='train, val image size (pixels)')
+    parser.add_argument('--rect', action='store_true', help='rectangular training')
+    parser.add_argument('--resume', nargs='?', const=True, default=False, help='resume most recent training')
+    parser.add_argument('--nosave', action='store_true', help='only save final checkpoint')
+    parser.add_argument('--noval', action='store_true', help='only validate final epoch')
+    parser.add_argument('--noautoanchor', action='store_true', help='disable AutoAnchor')
+    parser.add_argument('--evolve', type=int, nargs='?', const=300, help='evolve hyperparameters for x generations')
+    parser.add_argument('--bucket', type=str, default='', help='gsutil bucket')
+    parser.add_argument('--cache', type=str, nargs='?', const='ram', help='--cache images in "ram" (default) or "disk"')
+    parser.add_argument('--image-weights', action='store_true', help='use weighted image selection for training')
+    parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')
+    parser.add_argument('--multi-scale', action='store_true', help='vary img-size +/- 50%%')
+    parser.add_argument('--single-cls', action='store_true', help='train multi-class data as single-class')
+    parser.add_argument('--optimizer', type=str, choices=['SGD', 'Adam', 'AdamW'], default='SGD', help='optimizer')
+    parser.add_argument('--sync-bn', action='store_true', help='use SyncBatchNorm, only available in DDP mode')
+    parser.add_argument('--workers', type=int, default=8, help='max dataloader workers (per RANK in DDP mode)')
+    parser.add_argument('--project', default=ROOT / 'VOC2007_wm/train_sparity', help='save to project/name')
+    parser.add_argument('--name', default='exp', help='save to project/name')
+    parser.add_argument('--exist-ok', action='store_true', help='existing project/name ok, do not increment')
+    parser.add_argument('--quad', action='store_true', help='quad dataloader')
+    parser.add_argument('--cos-lr', action='store_true', help='cosine LR scheduler')
+    parser.add_argument('--label-smoothing', type=float, default=0.0, help='Label smoothing epsilon')
+    parser.add_argument('--patience', type=int, default=100, help='EarlyStopping patience (epochs without improvement)')
+    parser.add_argument('--freeze', nargs='+', type=int, default=[0], help='Freeze layers: backbone=10, first3=0 1 2')
+    parser.add_argument('--save-period', type=int, default=-1, help='Save checkpoint every x epochs (disabled if < 1)')
+    parser.add_argument('--local_rank', type=int, default=-1, help='DDP parameter, do not modify')
+
+    # Weights & Biases arguments
+    parser.add_argument('--entity', default=None, help='W&B: Entity')
+    parser.add_argument('--upload_dataset', nargs='?', const=True, default=False, help='W&B: Upload data, "val" option')
+    parser.add_argument('--bbox_interval', type=int, default=-1, help='W&B: Set bounding-box image logging interval')
+    parser.add_argument('--artifact_alias', type=str, default='latest', help='W&B: Version of dataset artifact to use')
+
+    opt = parser.parse_known_args()[0] if known else parser.parse_args()
+    return opt
+
+
+def main(opt, callbacks=Callbacks()):
+    # Checks
+    if RANK in [-1, 0]:
+        print_args(FILE.stem, opt)
+        check_git_status()
+        check_requirements(exclude=['thop'])
+
+    # Resume
+    if opt.resume and not check_wandb_resume(opt) and not opt.evolve:  # resume an interrupted run
+        ckpt = opt.resume if isinstance(opt.resume, str) else get_latest_run()  # specified or most recent path
+        assert os.path.isfile(ckpt), 'ERROR: --resume checkpoint does not exist'
+        with open(Path(ckpt).parent.parent / 'opt.yaml', errors='ignore') as f:
+            opt = argparse.Namespace(**yaml.safe_load(f))  # replace
+        opt.cfg, opt.weights, opt.resume = '', ckpt, True  # reinstate
+        LOGGER.info(f'Resuming training from {ckpt}')
+    else:
+        opt.data, opt.cfg, opt.hyp, opt.weights, opt.project = \
+            check_file(opt.data), check_yaml(opt.cfg), check_yaml(opt.hyp), str(opt.weights), str(opt.project)  # checks
+        assert len(opt.cfg) or len(opt.weights), 'either --cfg or --weights must be specified'
+        if opt.evolve:
+            if opt.project == str(ROOT / 'runs/train'):  # if default project name, rename to runs/evolve
+                opt.project = str(ROOT / 'runs/evolve')
+            opt.exist_ok, opt.resume = opt.resume, False  # pass resume to exist_ok and disable resume
+        opt.save_dir = str(increment_path(Path(opt.project) / opt.name, exist_ok=opt.exist_ok))
+
+    # DDP mode
+    device = select_device(opt.device, batch_size=opt.batch_size)
+    if LOCAL_RANK != -1:
+        msg = 'is not compatible with YOLOv5 Multi-GPU DDP training'
+        assert not opt.image_weights, f'--image-weights {msg}'
+        assert not opt.evolve, f'--evolve {msg}'
+        assert opt.batch_size != -1, f'AutoBatch with --batch-size -1 {msg}, please pass a valid --batch-size'
+        assert opt.batch_size % WORLD_SIZE == 0, f'--batch-size {opt.batch_size} must be multiple of WORLD_SIZE'
+        assert torch.cuda.device_count() > LOCAL_RANK, 'insufficient CUDA devices for DDP command'
+        torch.cuda.set_device(LOCAL_RANK)
+        device = torch.device('cuda', LOCAL_RANK)
+        dist.init_process_group(backend="nccl" if dist.is_nccl_available() else "gloo")
+
+    # Train
+    if not opt.evolve:
+        train(opt.hyp, opt, device, callbacks)
+        if WORLD_SIZE > 1 and RANK == 0:
+            LOGGER.info('Destroying process group... ')
+            dist.destroy_process_group()
+
+    # Evolve hyperparameters (optional)
+    else:
+        # Hyperparameter evolution metadata (mutation scale 0-1, lower_limit, upper_limit)
+        meta = {'lr0': (1, 1e-5, 1e-1),  # initial learning rate (SGD=1E-2, Adam=1E-3)
+                'lrf': (1, 0.01, 1.0),  # final OneCycleLR learning rate (lr0 * lrf)
+                'momentum': (0.3, 0.6, 0.98),  # SGD momentum/Adam beta1
+                'weight_decay': (1, 0.0, 0.001),  # optimizer weight decay
+                'warmup_epochs': (1, 0.0, 5.0),  # warmup epochs (fractions ok)
+                'warmup_momentum': (1, 0.0, 0.95),  # warmup initial momentum
+                'warmup_bias_lr': (1, 0.0, 0.2),  # warmup initial bias lr
+                'box': (1, 0.02, 0.2),  # box loss gain
+                'cls': (1, 0.2, 4.0),  # cls loss gain
+                'cls_pw': (1, 0.5, 2.0),  # cls BCELoss positive_weight
+                'obj': (1, 0.2, 4.0),  # obj loss gain (scale with pixels)
+                'obj_pw': (1, 0.5, 2.0),  # obj BCELoss positive_weight
+                'iou_t': (0, 0.1, 0.7),  # IoU training threshold
+                'anchor_t': (1, 2.0, 8.0),  # anchor-multiple threshold
+                'anchors': (2, 2.0, 10.0),  # anchors per output grid (0 to ignore)
+                'fl_gamma': (0, 0.0, 2.0),  # focal loss gamma (efficientDet default gamma=1.5)
+                'hsv_h': (1, 0.0, 0.1),  # image HSV-Hue augmentation (fraction)
+                'hsv_s': (1, 0.0, 0.9),  # image HSV-Saturation augmentation (fraction)
+                'hsv_v': (1, 0.0, 0.9),  # image HSV-Value augmentation (fraction)
+                'degrees': (1, 0.0, 45.0),  # image rotation (+/- deg)
+                'translate': (1, 0.0, 0.9),  # image translation (+/- fraction)
+                'scale': (1, 0.0, 0.9),  # image scale (+/- gain)
+                'shear': (1, 0.0, 10.0),  # image shear (+/- deg)
+                'perspective': (0, 0.0, 0.001),  # image perspective (+/- fraction), range 0-0.001
+                'flipud': (1, 0.0, 1.0),  # image flip up-down (probability)
+                'fliplr': (0, 0.0, 1.0),  # image flip left-right (probability)
+                'mosaic': (1, 0.0, 1.0),  # image mixup (probability)
+                'mixup': (1, 0.0, 1.0),  # image mixup (probability)
+                'copy_paste': (1, 0.0, 1.0)}  # segment copy-paste (probability)
+
+        with open(opt.hyp, errors='ignore') as f:
+            hyp = yaml.safe_load(f)  # load hyps dict
+            if 'anchors' not in hyp:  # anchors commented in hyp.yaml
+                hyp['anchors'] = 3
+        opt.noval, opt.nosave, save_dir = True, True, Path(opt.save_dir)  # only val/save final epoch
+        # ei = [isinstance(x, (int, float)) for x in hyp.values()]  # evolvable indices
+        evolve_yaml, evolve_csv = save_dir / 'hyp_evolve.yaml', save_dir / 'evolve.csv'
+        if opt.bucket:
+            os.system(f'gsutil cp gs://{opt.bucket}/evolve.csv {evolve_csv}')  # download evolve.csv if exists
+
+        for _ in range(opt.evolve):  # generations to evolve
+            if evolve_csv.exists():  # if evolve.csv exists: select best hyps and mutate
+                # Select parent(s)
+                parent = 'single'  # parent selection method: 'single' or 'weighted'
+                x = np.loadtxt(evolve_csv, ndmin=2, delimiter=',', skiprows=1)
+                n = min(5, len(x))  # number of previous results to consider
+                x = x[np.argsort(-fitness(x))][:n]  # top n mutations
+                w = fitness(x) - fitness(x).min() + 1E-6  # weights (sum > 0)
+                if parent == 'single' or len(x) == 1:
+                    # x = x[random.randint(0, n - 1)]  # random selection
+                    x = x[random.choices(range(n), weights=w)[0]]  # weighted selection
+                elif parent == 'weighted':
+                    x = (x * w.reshape(n, 1)).sum(0) / w.sum()  # weighted combination
+
+                # Mutate
+                mp, s = 0.8, 0.2  # mutation probability, sigma
+                npr = np.random
+                npr.seed(int(time.time()))
+                g = np.array([meta[k][0] for k in hyp.keys()])  # gains 0-1
+                ng = len(meta)
+                v = np.ones(ng)
+                while all(v == 1):  # mutate until a change occurs (prevent duplicates)
+                    v = (g * (npr.random(ng) < mp) * npr.randn(ng) * npr.random() * s + 1).clip(0.3, 3.0)
+                for i, k in enumerate(hyp.keys()):  # plt.hist(v.ravel(), 300)
+                    hyp[k] = float(x[i + 7] * v[i])  # mutate
+
+            # Constrain to limits
+            for k, v in meta.items():
+                hyp[k] = max(hyp[k], v[1])  # lower limit
+                hyp[k] = min(hyp[k], v[2])  # upper limit
+                hyp[k] = round(hyp[k], 5)  # significant digits
+
+            # Train mutation
+            results = train(hyp.copy(), opt, device, callbacks)
+            callbacks = Callbacks()
+            # Write mutation results
+            print_mutation(results, hyp.copy(), save_dir, opt.bucket)
+
+        # Plot results
+        plot_evolve(evolve_csv)
+        LOGGER.info(f'Hyperparameter evolution finished {opt.evolve} generations\n'
+                    f"Results saved to {colorstr('bold', save_dir)}\n"
+                    f'Usage example: $ python train.py --hyp {evolve_yaml}')
+
+
+def run(**kwargs):
+    # Usage: import train; train.run(data='coco128.yaml', imgsz=320, weights='yolov5m.pt')
+    opt = parse_opt(True)
+    for k, v in kwargs.items():
+        setattr(opt, k, v)
+    main(opt)
+    return opt
+
+
+if __name__ == "__main__":
+    opt = parse_opt()
+    main(opt)

utils/__init__.py

@@ -0,0 +1,37 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+"""
+utils/initialization
+"""
+
+
+def notebook_init(verbose=True):
+    # Check system software and hardware
+    print('Checking setup...')
+
+    import os
+    import shutil
+
+    from utils.general import check_requirements, emojis, is_colab
+    from utils.torch_utils import select_device  # imports
+
+    check_requirements(('psutil', 'IPython'))
+    import psutil
+    from IPython import display  # to display images and clear console output
+
+    if is_colab():
+        shutil.rmtree('/content/sample_data', ignore_errors=True)  # remove colab /sample_data directory
+
+    if verbose:
+        # System info
+        # gb = 1 / 1000 ** 3  # bytes to GB
+        gib = 1 / 1024 ** 3  # bytes to GiB
+        ram = psutil.virtual_memory().total
+        total, used, free = shutil.disk_usage("/")
+        display.clear_output()
+        s = f'({os.cpu_count()} CPUs, {ram * gib:.1f} GB RAM, {(total - free) * gib:.1f}/{total * gib:.1f} GB disk)'
+    else:
+        s = ''
+
+    select_device(newline=False)
+    print(emojis(f'Setup complete ✅ {s}'))
+    return display

utils/activations.py

@@ -0,0 +1,101 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+"""
+Activation functions
+"""
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+
+# SiLU https://arxiv.org/pdf/1606.08415.pdf ----------------------------------------------------------------------------
+class SiLU(nn.Module):  # export-friendly version of nn.SiLU()
+    @staticmethod
+    def forward(x):
+        return x * torch.sigmoid(x)
+
+
+class Hardswish(nn.Module):  # export-friendly version of nn.Hardswish()
+    @staticmethod
+    def forward(x):
+        # return x * F.hardsigmoid(x)  # for TorchScript and CoreML
+        return x * F.hardtanh(x + 3, 0.0, 6.0) / 6.0  # for TorchScript, CoreML and ONNX
+
+
+# Mish https://github.com/digantamisra98/Mish --------------------------------------------------------------------------
+class Mish(nn.Module):
+    @staticmethod
+    def forward(x):
+        return x * F.softplus(x).tanh()
+
+
+class MemoryEfficientMish(nn.Module):
+    class F(torch.autograd.Function):
+        @staticmethod
+        def forward(ctx, x):
+            ctx.save_for_backward(x)
+            return x.mul(torch.tanh(F.softplus(x)))  # x * tanh(ln(1 + exp(x)))
+
+        @staticmethod
+        def backward(ctx, grad_output):
+            x = ctx.saved_tensors[0]
+            sx = torch.sigmoid(x)
+            fx = F.softplus(x).tanh()
+            return grad_output * (fx + x * sx * (1 - fx * fx))
+
+    def forward(self, x):
+        return self.F.apply(x)
+
+
+# FReLU https://arxiv.org/abs/2007.11824 -------------------------------------------------------------------------------
+class FReLU(nn.Module):
+    def __init__(self, c1, k=3):  # ch_in, kernel
+        super().__init__()
+        self.conv = nn.Conv2d(c1, c1, k, 1, 1, groups=c1, bias=False)
+        self.bn = nn.BatchNorm2d(c1)
+
+    def forward(self, x):
+        return torch.max(x, self.bn(self.conv(x)))
+
+
+# ACON https://arxiv.org/pdf/2009.04759.pdf ----------------------------------------------------------------------------
+class AconC(nn.Module):
+    r""" ACON activation (activate or not).
+    AconC: (p1*x-p2*x) * sigmoid(beta*(p1*x-p2*x)) + p2*x, beta is a learnable parameter
+    according to "Activate or Not: Learning Customized Activation" <https://arxiv.org/pdf/2009.04759.pdf>.
+    """
+
+    def __init__(self, c1):
+        super().__init__()
+        self.p1 = nn.Parameter(torch.randn(1, c1, 1, 1))
+        self.p2 = nn.Parameter(torch.randn(1, c1, 1, 1))
+        self.beta = nn.Parameter(torch.ones(1, c1, 1, 1))
+
+    def forward(self, x):
+        dpx = (self.p1 - self.p2) * x
+        return dpx * torch.sigmoid(self.beta * dpx) + self.p2 * x
+
+
+class MetaAconC(nn.Module):
+    r""" ACON activation (activate or not).
+    MetaAconC: (p1*x-p2*x) * sigmoid(beta*(p1*x-p2*x)) + p2*x, beta is generated by a small network
+    according to "Activate or Not: Learning Customized Activation" <https://arxiv.org/pdf/2009.04759.pdf>.
+    """
+
+    def __init__(self, c1, k=1, s=1, r=16):  # ch_in, kernel, stride, r
+        super().__init__()
+        c2 = max(r, c1 // r)
+        self.p1 = nn.Parameter(torch.randn(1, c1, 1, 1))
+        self.p2 = nn.Parameter(torch.randn(1, c1, 1, 1))
+        self.fc1 = nn.Conv2d(c1, c2, k, s, bias=True)
+        self.fc2 = nn.Conv2d(c2, c1, k, s, bias=True)
+        # self.bn1 = nn.BatchNorm2d(c2)
+        # self.bn2 = nn.BatchNorm2d(c1)
+
+    def forward(self, x):
+        y = x.mean(dim=2, keepdims=True).mean(dim=3, keepdims=True)
+        # batch-size 1 bug/instabilities https://github.com/ultralytics/yolov5/issues/2891
+        # beta = torch.sigmoid(self.bn2(self.fc2(self.bn1(self.fc1(y)))))  # bug/unstable
+        beta = torch.sigmoid(self.fc2(self.fc1(y)))  # bug patch BN layers removed
+        dpx = (self.p1 - self.p2) * x
+        return dpx * torch.sigmoid(beta * dpx) + self.p2 * x

utils/augmentations.py

@@ -0,0 +1,277 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+"""
+Image augmentation functions
+"""
+
+import math
+import random
+
+import cv2
+import numpy as np
+
+from utils.general import LOGGER, check_version, colorstr, resample_segments, segment2box
+from utils.metrics import bbox_ioa
+
+
+class Albumentations:
+    # YOLOv5 Albumentations class (optional, only used if package is installed)
+    def __init__(self):
+        self.transform = None
+        try:
+            import albumentations as A
+            check_version(A.__version__, '1.0.3', hard=True)  # version requirement
+
+            self.transform = A.Compose([
+                A.Blur(p=0.01),
+                A.MedianBlur(p=0.01),
+                A.ToGray(p=0.01),
+                A.CLAHE(p=0.01),
+                A.RandomBrightnessContrast(p=0.0),
+                A.RandomGamma(p=0.0),
+                A.ImageCompression(quality_lower=75, p=0.0)],
+                bbox_params=A.BboxParams(format='yolo', label_fields=['class_labels']))
+
+            LOGGER.info(colorstr('albumentations: ') + ', '.join(f'{x}' for x in self.transform.transforms if x.p))
+        except ImportError:  # package not installed, skip
+            pass
+        except Exception as e:
+            LOGGER.info(colorstr('albumentations: ') + f'{e}')
+
+    def __call__(self, im, labels, p=1.0):
+        if self.transform and random.random() < p:
+            new = self.transform(image=im, bboxes=labels[:, 1:], class_labels=labels[:, 0])  # transformed
+            im, labels = new['image'], np.array([[c, *b] for c, b in zip(new['class_labels'], new['bboxes'])])
+        return im, labels
+
+
+def augment_hsv(im, hgain=0.5, sgain=0.5, vgain=0.5):
+    # HSV color-space augmentation
+    if hgain or sgain or vgain:
+        r = np.random.uniform(-1, 1, 3) * [hgain, sgain, vgain] + 1  # random gains
+        hue, sat, val = cv2.split(cv2.cvtColor(im, cv2.COLOR_BGR2HSV))
+        dtype = im.dtype  # uint8
+
+        x = np.arange(0, 256, dtype=r.dtype)
+        lut_hue = ((x * r[0]) % 180).astype(dtype)
+        lut_sat = np.clip(x * r[1], 0, 255).astype(dtype)
+        lut_val = np.clip(x * r[2], 0, 255).astype(dtype)
+
+        im_hsv = cv2.merge((cv2.LUT(hue, lut_hue), cv2.LUT(sat, lut_sat), cv2.LUT(val, lut_val)))
+        cv2.cvtColor(im_hsv, cv2.COLOR_HSV2BGR, dst=im)  # no return needed
+
+
+def hist_equalize(im, clahe=True, bgr=False):
+    # Equalize histogram on BGR image 'im' with im.shape(n,m,3) and range 0-255
+    yuv = cv2.cvtColor(im, cv2.COLOR_BGR2YUV if bgr else cv2.COLOR_RGB2YUV)
+    if clahe:
+        c = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8, 8))
+        yuv[:, :, 0] = c.apply(yuv[:, :, 0])
+    else:
+        yuv[:, :, 0] = cv2.equalizeHist(yuv[:, :, 0])  # equalize Y channel histogram
+    return cv2.cvtColor(yuv, cv2.COLOR_YUV2BGR if bgr else cv2.COLOR_YUV2RGB)  # convert YUV image to RGB
+
+
+def replicate(im, labels):
+    # Replicate labels
+    h, w = im.shape[:2]
+    boxes = labels[:, 1:].astype(int)
+    x1, y1, x2, y2 = boxes.T
+    s = ((x2 - x1) + (y2 - y1)) / 2  # side length (pixels)
+    for i in s.argsort()[:round(s.size * 0.5)]:  # smallest indices
+        x1b, y1b, x2b, y2b = boxes[i]
+        bh, bw = y2b - y1b, x2b - x1b
+        yc, xc = int(random.uniform(0, h - bh)), int(random.uniform(0, w - bw))  # offset x, y
+        x1a, y1a, x2a, y2a = [xc, yc, xc + bw, yc + bh]
+        im[y1a:y2a, x1a:x2a] = im[y1b:y2b, x1b:x2b]  # im4[ymin:ymax, xmin:xmax]
+        labels = np.append(labels, [[labels[i, 0], x1a, y1a, x2a, y2a]], axis=0)
+
+    return im, labels
+
+
+def letterbox(im, new_shape=(640, 640), color=(114, 114, 114), auto=True, scaleFill=False, scaleup=True, stride=32):
+    # Resize and pad image while meeting stride-multiple constraints
+    shape = im.shape[:2]  # current shape [height, width]
+    if isinstance(new_shape, int):
+        new_shape = (new_shape, new_shape)
+
+    # Scale ratio (new / old)
+    r = min(new_shape[0] / shape[0], new_shape[1] / shape[1])
+    if not scaleup:  # only scale down, do not scale up (for better val mAP)
+        r = min(r, 1.0)
+
+    # Compute padding
+    ratio = r, r  # width, height ratios
+    new_unpad = int(round(shape[1] * r)), int(round(shape[0] * r))
+    dw, dh = new_shape[1] - new_unpad[0], new_shape[0] - new_unpad[1]  # wh padding
+    if auto:  # minimum rectangle
+        dw, dh = np.mod(dw, stride), np.mod(dh, stride)  # wh padding
+    elif scaleFill:  # stretch
+        dw, dh = 0.0, 0.0
+        new_unpad = (new_shape[1], new_shape[0])
+        ratio = new_shape[1] / shape[1], new_shape[0] / shape[0]  # width, height ratios
+
+    dw /= 2  # divide padding into 2 sides
+    dh /= 2
+
+    if shape[::-1] != new_unpad:  # resize
+        im = cv2.resize(im, new_unpad, interpolation=cv2.INTER_LINEAR)
+    top, bottom = int(round(dh - 0.1)), int(round(dh + 0.1))
+    left, right = int(round(dw - 0.1)), int(round(dw + 0.1))
+    im = cv2.copyMakeBorder(im, top, bottom, left, right, cv2.BORDER_CONSTANT, value=color)  # add border
+    return im, ratio, (dw, dh)
+
+
+def random_perspective(im, targets=(), segments=(), degrees=10, translate=.1, scale=.1, shear=10, perspective=0.0,
+                       border=(0, 0)):
+    # torchvision.transforms.RandomAffine(degrees=(-10, 10), translate=(0.1, 0.1), scale=(0.9, 1.1), shear=(-10, 10))
+    # targets = [cls, xyxy]
+
+    height = im.shape[0] + border[0] * 2  # shape(h,w,c)
+    width = im.shape[1] + border[1] * 2
+
+    # Center
+    C = np.eye(3)
+    C[0, 2] = -im.shape[1] / 2  # x translation (pixels)
+    C[1, 2] = -im.shape[0] / 2  # y translation (pixels)
+
+    # Perspective
+    P = np.eye(3)
+    P[2, 0] = random.uniform(-perspective, perspective)  # x perspective (about y)
+    P[2, 1] = random.uniform(-perspective, perspective)  # y perspective (about x)
+
+    # Rotation and Scale
+    R = np.eye(3)
+    a = random.uniform(-degrees, degrees)
+    # a += random.choice([-180, -90, 0, 90])  # add 90deg rotations to small rotations
+    s = random.uniform(1 - scale, 1 + scale)
+    # s = 2 ** random.uniform(-scale, scale)
+    R[:2] = cv2.getRotationMatrix2D(angle=a, center=(0, 0), scale=s)
+
+    # Shear
+    S = np.eye(3)
+    S[0, 1] = math.tan(random.uniform(-shear, shear) * math.pi / 180)  # x shear (deg)
+    S[1, 0] = math.tan(random.uniform(-shear, shear) * math.pi / 180)  # y shear (deg)
+
+    # Translation
+    T = np.eye(3)
+    T[0, 2] = random.uniform(0.5 - translate, 0.5 + translate) * width  # x translation (pixels)
+    T[1, 2] = random.uniform(0.5 - translate, 0.5 + translate) * height  # y translation (pixels)
+
+    # Combined rotation matrix
+    M = T @ S @ R @ P @ C  # order of operations (right to left) is IMPORTANT
+    if (border[0] != 0) or (border[1] != 0) or (M != np.eye(3)).any():  # image changed
+        if perspective:
+            im = cv2.warpPerspective(im, M, dsize=(width, height), borderValue=(114, 114, 114))
+        else:  # affine
+            im = cv2.warpAffine(im, M[:2], dsize=(width, height), borderValue=(114, 114, 114))
+
+    # Visualize
+    # import matplotlib.pyplot as plt
+    # ax = plt.subplots(1, 2, figsize=(12, 6))[1].ravel()
+    # ax[0].imshow(im[:, :, ::-1])  # base
+    # ax[1].imshow(im2[:, :, ::-1])  # warped
+
+    # Transform label coordinates
+    n = len(targets)
+    if n:
+        use_segments = any(x.any() for x in segments)
+        new = np.zeros((n, 4))
+        if use_segments:  # warp segments
+            segments = resample_segments(segments)  # upsample
+            for i, segment in enumerate(segments):
+                xy = np.ones((len(segment), 3))
+                xy[:, :2] = segment
+                xy = xy @ M.T  # transform
+                xy = xy[:, :2] / xy[:, 2:3] if perspective else xy[:, :2]  # perspective rescale or affine
+
+                # clip
+                new[i] = segment2box(xy, width, height)
+
+        else:  # warp boxes
+            xy = np.ones((n * 4, 3))
+            xy[:, :2] = targets[:, [1, 2, 3, 4, 1, 4, 3, 2]].reshape(n * 4, 2)  # x1y1, x2y2, x1y2, x2y1
+            xy = xy @ M.T  # transform
+            xy = (xy[:, :2] / xy[:, 2:3] if perspective else xy[:, :2]).reshape(n, 8)  # perspective rescale or affine
+
+            # create new boxes
+            x = xy[:, [0, 2, 4, 6]]
+            y = xy[:, [1, 3, 5, 7]]
+            new = np.concatenate((x.min(1), y.min(1), x.max(1), y.max(1))).reshape(4, n).T
+
+            # clip
+            new[:, [0, 2]] = new[:, [0, 2]].clip(0, width)
+            new[:, [1, 3]] = new[:, [1, 3]].clip(0, height)
+
+        # filter candidates
+        i = box_candidates(box1=targets[:, 1:5].T * s, box2=new.T, area_thr=0.01 if use_segments else 0.10)
+        targets = targets[i]
+        targets[:, 1:5] = new[i]
+
+    return im, targets
+
+
+def copy_paste(im, labels, segments, p=0.5):
+    # Implement Copy-Paste augmentation https://arxiv.org/abs/2012.07177, labels as nx5 np.array(cls, xyxy)
+    n = len(segments)
+    if p and n:
+        h, w, c = im.shape  # height, width, channels
+        im_new = np.zeros(im.shape, np.uint8)
+        for j in random.sample(range(n), k=round(p * n)):
+            l, s = labels[j], segments[j]
+            box = w - l[3], l[2], w - l[1], l[4]
+            ioa = bbox_ioa(box, labels[:, 1:5])  # intersection over area
+            if (ioa < 0.30).all():  # allow 30% obscuration of existing labels
+                labels = np.concatenate((labels, [[l[0], *box]]), 0)
+                segments.append(np.concatenate((w - s[:, 0:1], s[:, 1:2]), 1))
+                cv2.drawContours(im_new, [segments[j].astype(np.int32)], -1, (255, 255, 255), cv2.FILLED)
+
+        result = cv2.bitwise_and(src1=im, src2=im_new)
+        result = cv2.flip(result, 1)  # augment segments (flip left-right)
+        i = result > 0  # pixels to replace
+        # i[:, :] = result.max(2).reshape(h, w, 1)  # act over ch
+        im[i] = result[i]  # cv2.imwrite('debug.jpg', im)  # debug
+
+    return im, labels, segments
+
+
+def cutout(im, labels, p=0.5):
+    # Applies image cutout augmentation https://arxiv.org/abs/1708.04552
+    if random.random() < p:
+        h, w = im.shape[:2]
+        scales = [0.5] * 1 + [0.25] * 2 + [0.125] * 4 + [0.0625] * 8 + [0.03125] * 16  # image size fraction
+        for s in scales:
+            mask_h = random.randint(1, int(h * s))  # create random masks
+            mask_w = random.randint(1, int(w * s))
+
+            # box
+            xmin = max(0, random.randint(0, w) - mask_w // 2)
+            ymin = max(0, random.randint(0, h) - mask_h // 2)
+            xmax = min(w, xmin + mask_w)
+            ymax = min(h, ymin + mask_h)
+
+            # apply random color mask
+            im[ymin:ymax, xmin:xmax] = [random.randint(64, 191) for _ in range(3)]
+
+            # return unobscured labels
+            if len(labels) and s > 0.03:
+                box = np.array([xmin, ymin, xmax, ymax], dtype=np.float32)
+                ioa = bbox_ioa(box, labels[:, 1:5])  # intersection over area
+                labels = labels[ioa < 0.60]  # remove >60% obscured labels
+
+    return labels
+
+
+def mixup(im, labels, im2, labels2):
+    # Applies MixUp augmentation https://arxiv.org/pdf/1710.09412.pdf
+    r = np.random.beta(32.0, 32.0)  # mixup ratio, alpha=beta=32.0
+    im = (im * r + im2 * (1 - r)).astype(np.uint8)
+    labels = np.concatenate((labels, labels2), 0)
+    return im, labels
+
+
+def box_candidates(box1, box2, wh_thr=2, ar_thr=100, area_thr=0.1, eps=1e-16):  # box1(4,n), box2(4,n)
+    # Compute candidate boxes: box1 before augment, box2 after augment, wh_thr (pixels), aspect_ratio_thr, area_ratio
+    w1, h1 = box1[2] - box1[0], box1[3] - box1[1]
+    w2, h2 = box2[2] - box2[0], box2[3] - box2[1]
+    ar = np.maximum(w2 / (h2 + eps), h2 / (w2 + eps))  # aspect ratio
+    return (w2 > wh_thr) & (h2 > wh_thr) & (w2 * h2 / (w1 * h1 + eps) > area_thr) & (ar < ar_thr)  # candidates

utils/autoanchor.py

@@ -0,0 +1,165 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+"""
+AutoAnchor utils
+"""
+
+import random
+
+import numpy as np
+import torch
+import yaml
+from tqdm import tqdm
+
+from utils.general import LOGGER, colorstr, emojis
+
+PREFIX = colorstr('AutoAnchor: ')
+
+
+def check_anchor_order(m):
+    # Check anchor order against stride order for YOLOv5 Detect() module m, and correct if necessary
+    a = m.anchors.prod(-1).view(-1)  # anchor area
+    da = a[-1] - a[0]  # delta a
+    ds = m.stride[-1] - m.stride[0]  # delta s
+    if da.sign() != ds.sign():  # same order
+        LOGGER.info(f'{PREFIX}Reversing anchor order')
+        m.anchors[:] = m.anchors.flip(0)
+
+
+def check_anchors(dataset, model, thr=4.0, imgsz=640):
+    # Check anchor fit to data, recompute if necessary
+    m = model.module.model[-1] if hasattr(model, 'module') else model.model[-1]  # Detect()
+    shapes = imgsz * dataset.shapes / dataset.shapes.max(1, keepdims=True)
+    scale = np.random.uniform(0.9, 1.1, size=(shapes.shape[0], 1))  # augment scale
+    wh = torch.tensor(np.concatenate([l[:, 3:5] * s for s, l in zip(shapes * scale, dataset.labels)])).float()  # wh
+
+    def metric(k):  # compute metric
+        r = wh[:, None] / k[None]
+        x = torch.min(r, 1 / r).min(2)[0]  # ratio metric
+        best = x.max(1)[0]  # best_x
+        aat = (x > 1 / thr).float().sum(1).mean()  # anchors above threshold
+        bpr = (best > 1 / thr).float().mean()  # best possible recall
+        return bpr, aat
+
+    anchors = m.anchors.clone() * m.stride.to(m.anchors.device).view(-1, 1, 1)  # current anchors
+    bpr, aat = metric(anchors.cpu().view(-1, 2))
+    s = f'\n{PREFIX}{aat:.2f} anchors/target, {bpr:.3f} Best Possible Recall (BPR). '
+    if bpr > 0.98:  # threshold to recompute
+        LOGGER.info(emojis(f'{s}Current anchors are a good fit to dataset ✅'))
+    else:
+        LOGGER.info(emojis(f'{s}Anchors are a poor fit to dataset ⚠️, attempting to improve...'))
+        na = m.anchors.numel() // 2  # number of anchors
+        try:
+            anchors = kmean_anchors(dataset, n=na, img_size=imgsz, thr=thr, gen=1000, verbose=False)
+        except Exception as e:
+            LOGGER.info(f'{PREFIX}ERROR: {e}')
+        new_bpr = metric(anchors)[0]
+        if new_bpr > bpr:  # replace anchors
+            anchors = torch.tensor(anchors, device=m.anchors.device).type_as(m.anchors)
+            m.anchors[:] = anchors.clone().view_as(m.anchors) / m.stride.to(m.anchors.device).view(-1, 1, 1)  # loss
+            check_anchor_order(m)
+            LOGGER.info(f'{PREFIX}New anchors saved to model. Update model *.yaml to use these anchors in the future.')
+        else:
+            LOGGER.info(f'{PREFIX}Original anchors better than new anchors. Proceeding with original anchors.')
+
+
+def kmean_anchors(dataset='./data/coco128.yaml', n=9, img_size=640, thr=4.0, gen=1000, verbose=True):
+    """ Creates kmeans-evolved anchors from training dataset
+
+        Arguments:
+            dataset: path to data.yaml, or a loaded dataset
+            n: number of anchors
+            img_size: image size used for training
+            thr: anchor-label wh ratio threshold hyperparameter hyp['anchor_t'] used for training, default=4.0
+            gen: generations to evolve anchors using genetic algorithm
+            verbose: print all results
+
+        Return:
+            k: kmeans evolved anchors
+
+        Usage:
+            from utils.autoanchor import *; _ = kmean_anchors()
+    """
+    from scipy.cluster.vq import kmeans
+
+    npr = np.random
+    thr = 1 / thr
+
+    def metric(k, wh):  # compute metrics
+        r = wh[:, None] / k[None]
+        x = torch.min(r, 1 / r).min(2)[0]  # ratio metric
+        # x = wh_iou(wh, torch.tensor(k))  # iou metric
+        return x, x.max(1)[0]  # x, best_x
+
+    def anchor_fitness(k):  # mutation fitness
+        _, best = metric(torch.tensor(k, dtype=torch.float32), wh)
+        return (best * (best > thr).float()).mean()  # fitness
+
+    def print_results(k, verbose=True):
+        k = k[np.argsort(k.prod(1))]  # sort small to large
+        x, best = metric(k, wh0)
+        bpr, aat = (best > thr).float().mean(), (x > thr).float().mean() * n  # best possible recall, anch > thr
+        s = f'{PREFIX}thr={thr:.2f}: {bpr:.4f} best possible recall, {aat:.2f} anchors past thr\n' \
+            f'{PREFIX}n={n}, img_size={img_size}, metric_all={x.mean():.3f}/{best.mean():.3f}-mean/best, ' \
+            f'past_thr={x[x > thr].mean():.3f}-mean: '
+        for i, x in enumerate(k):
+            s += '%i,%i, ' % (round(x[0]), round(x[1]))
+        if verbose:
+            LOGGER.info(s[:-2])
+        return k
+
+    if isinstance(dataset, str):  # *.yaml file
+        with open(dataset, errors='ignore') as f:
+            data_dict = yaml.safe_load(f)  # model dict
+        from utils.datasets import LoadImagesAndLabels
+        dataset = LoadImagesAndLabels(data_dict['train'], augment=True, rect=True)
+
+    # Get label wh
+    shapes = img_size * dataset.shapes / dataset.shapes.max(1, keepdims=True)
+    wh0 = np.concatenate([l[:, 3:5] * s for s, l in zip(shapes, dataset.labels)])  # wh
+
+    # Filter
+    i = (wh0 < 3.0).any(1).sum()
+    if i:
+        LOGGER.info(f'{PREFIX}WARNING: Extremely small objects found. {i} of {len(wh0)} labels are < 3 pixels in size.')
+    wh = wh0[(wh0 >= 2.0).any(1)]  # filter > 2 pixels
+    # wh = wh * (npr.rand(wh.shape[0], 1) * 0.9 + 0.1)  # multiply by random scale 0-1
+
+    # Kmeans calculation
+    LOGGER.info(f'{PREFIX}Running kmeans for {n} anchors on {len(wh)} points...')
+    s = wh.std(0)  # sigmas for whitening
+    k = kmeans(wh / s, n, iter=30)[0] * s  # points
+    if len(k) != n:  # kmeans may return fewer points than requested if wh is insufficient or too similar
+        LOGGER.warning(f'{PREFIX}WARNING: scipy.cluster.vq.kmeans returned only {len(k)} of {n} requested points')
+        k = np.sort(npr.rand(n * 2)).reshape(n, 2) * img_size  # random init
+    wh = torch.tensor(wh, dtype=torch.float32)  # filtered
+    wh0 = torch.tensor(wh0, dtype=torch.float32)  # unfiltered
+    k = print_results(k, verbose=False)
+
+    # Plot
+    # k, d = [None] * 20, [None] * 20
+    # for i in tqdm(range(1, 21)):
+    #     k[i-1], d[i-1] = kmeans(wh / s, i)  # points, mean distance
+    # fig, ax = plt.subplots(1, 2, figsize=(14, 7), tight_layout=True)
+    # ax = ax.ravel()
+    # ax[0].plot(np.arange(1, 21), np.array(d) ** 2, marker='.')
+    # fig, ax = plt.subplots(1, 2, figsize=(14, 7))  # plot wh
+    # ax[0].hist(wh[wh[:, 0]<100, 0],400)
+    # ax[1].hist(wh[wh[:, 1]<100, 1],400)
+    # fig.savefig('wh.png', dpi=200)
+
+    # Evolve
+    f, sh, mp, s = anchor_fitness(k), k.shape, 0.9, 0.1  # fitness, generations, mutation prob, sigma
+    pbar = tqdm(range(gen), desc=f'{PREFIX}Evolving anchors with Genetic Algorithm:')  # progress bar
+    for _ in pbar:
+        v = np.ones(sh)
+        while (v == 1).all():  # mutate until a change occurs (prevent duplicates)
+            v = ((npr.random(sh) < mp) * random.random() * npr.randn(*sh) * s + 1).clip(0.3, 3.0)
+        kg = (k.copy() * v).clip(min=2.0)
+        fg = anchor_fitness(kg)
+        if fg > f:
+            f, k = fg, kg.copy()
+            pbar.desc = f'{PREFIX}Evolving anchors with Genetic Algorithm: fitness = {f:.4f}'
+            if verbose:
+                print_results(k, verbose)
+
+    return print_results(k)

utils/autobatch.py

@@ -0,0 +1,57 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+"""
+Auto-batch utils
+"""
+
+from copy import deepcopy
+
+import numpy as np
+import torch
+from torch.cuda import amp
+
+from utils.general import LOGGER, colorstr
+from utils.torch_utils import profile
+
+
+def check_train_batch_size(model, imgsz=640):
+    # Check YOLOv5 training batch size
+    with amp.autocast():
+        return autobatch(deepcopy(model).train(), imgsz)  # compute optimal batch size
+
+
+def autobatch(model, imgsz=640, fraction=0.9, batch_size=16):
+    # Automatically estimate best batch size to use `fraction` of available CUDA memory
+    # Usage:
+    #     import torch
+    #     from utils.autobatch import autobatch
+    #     model = torch.hub.load('ultralytics/yolov5', 'yolov5s', autoshape=False)
+    #     print(autobatch(model))
+
+    prefix = colorstr('AutoBatch: ')
+    LOGGER.info(f'{prefix}Computing optimal batch size for --imgsz {imgsz}')
+    device = next(model.parameters()).device  # get model device
+    if device.type == 'cpu':
+        LOGGER.info(f'{prefix}CUDA not detected, using default CPU batch-size {batch_size}')
+        return batch_size
+
+    d = str(device).upper()  # 'CUDA:0'
+    properties = torch.cuda.get_device_properties(device)  # device properties
+    t = properties.total_memory / 1024 ** 3  # (GiB)
+    r = torch.cuda.memory_reserved(device) / 1024 ** 3  # (GiB)
+    a = torch.cuda.memory_allocated(device) / 1024 ** 3  # (GiB)
+    f = t - (r + a)  # free inside reserved
+    LOGGER.info(f'{prefix}{d} ({properties.name}) {t:.2f}G total, {r:.2f}G reserved, {a:.2f}G allocated, {f:.2f}G free')
+
+    batch_sizes = [1, 2, 4, 8, 16]
+    try:
+        img = [torch.zeros(b, 3, imgsz, imgsz) for b in batch_sizes]
+        y = profile(img, model, n=3, device=device)
+    except Exception as e:
+        LOGGER.warning(f'{prefix}{e}')
+
+    y = [x[2] for x in y if x]  # memory [2]
+    batch_sizes = batch_sizes[:len(y)]
+    p = np.polyfit(batch_sizes, y, deg=1)  # first degree polynomial fit
+    b = int((f * fraction - p[1]) / p[0])  # y intercept (optimal batch size)
+    LOGGER.info(f'{prefix}Using batch-size {b} for {d} {t * fraction:.2f}G/{t:.2f}G ({fraction * 100:.0f}%)')
+    return b

utils/aws/mime.sh

@@ -0,0 +1,26 @@
+# AWS EC2 instance startup 'MIME' script https://aws.amazon.com/premiumsupport/knowledge-center/execute-user-data-ec2/
+# This script will run on every instance restart, not only on first start
+# --- DO NOT COPY ABOVE COMMENTS WHEN PASTING INTO USERDATA ---
+
+Content-Type: multipart/mixed; boundary="//"
+MIME-Version: 1.0
+
+--//
+Content-Type: text/cloud-config; charset="us-ascii"
+MIME-Version: 1.0
+Content-Transfer-Encoding: 7bit
+Content-Disposition: attachment; filename="cloud-config.txt"
+
+#cloud-config
+cloud_final_modules:
+- [scripts-user, always]
+
+--//
+Content-Type: text/x-shellscript; charset="us-ascii"
+MIME-Version: 1.0
+Content-Transfer-Encoding: 7bit
+Content-Disposition: attachment; filename="userdata.txt"
+
+#!/bin/bash
+# --- paste contents of userdata.sh here ---
+--//

utils/aws/resume.py

@@ -0,0 +1,40 @@
+# Resume all interrupted trainings in yolov5/ dir including DDP trainings
+# Usage: $ python utils/aws/resume.py
+
+import os
+import sys
+from pathlib import Path
+
+import torch
+import yaml
+
+FILE = Path(__file__).resolve()
+ROOT = FILE.parents[2]  # YOLOv5 root directory
+if str(ROOT) not in sys.path:
+    sys.path.append(str(ROOT))  # add ROOT to PATH
+
+port = 0  # --master_port
+path = Path('').resolve()
+for last in path.rglob('*/**/last.pt'):
+    ckpt = torch.load(last)
+    if ckpt['optimizer'] is None:
+        continue
+
+    # Load opt.yaml
+    with open(last.parent.parent / 'opt.yaml', errors='ignore') as f:
+        opt = yaml.safe_load(f)
+
+    # Get device count
+    d = opt['device'].split(',')  # devices
+    nd = len(d)  # number of devices
+    ddp = nd > 1 or (nd == 0 and torch.cuda.device_count() > 1)  # distributed data parallel
+
+    if ddp:  # multi-GPU
+        port += 1
+        cmd = f'python -m torch.distributed.run --nproc_per_node {nd} --master_port {port} train.py --resume {last}'
+    else:  # single-GPU
+        cmd = f'python train.py --resume {last}'
+
+    cmd += ' > /dev/null 2>&1 &'  # redirect output to dev/null and run in daemon thread
+    print(cmd)
+    os.system(cmd)

utils/aws/userdata.sh

@@ -0,0 +1,27 @@
+#!/bin/bash
+# AWS EC2 instance startup script https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/user-data.html
+# This script will run only once on first instance start (for a re-start script see mime.sh)
+# /home/ubuntu (ubuntu) or /home/ec2-user (amazon-linux) is working dir
+# Use >300 GB SSD
+
+cd home/ubuntu
+if [ ! -d yolov5 ]; then
+  echo "Running first-time script." # install dependencies, download COCO, pull Docker
+  git clone https://github.com/ultralytics/yolov5 -b master && sudo chmod -R 777 yolov5
+  cd yolov5
+  bash data/scripts/get_coco.sh && echo "COCO done." &
+  sudo docker pull ultralytics/yolov5:latest && echo "Docker done." &
+  python -m pip install --upgrade pip && pip install -r requirements.txt && python detect.py && echo "Requirements done." &
+  wait && echo "All tasks done." # finish background tasks
+else
+  echo "Running re-start script." # resume interrupted runs
+  i=0
+  list=$(sudo docker ps -qa) # container list i.e. $'one\ntwo\nthree\nfour'
+  while IFS= read -r id; do
+    ((i++))
+    echo "restarting container $i: $id"
+    sudo docker start $id
+    # sudo docker exec -it $id python train.py --resume # single-GPU
+    sudo docker exec -d $id python utils/aws/resume.py # multi-scenario
+  done <<<"$list"
+fi

utils/benchmarks.py

@@ -0,0 +1,92 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+"""
+Run YOLOv5 benchmarks on all supported export formats
+
+Format                      | `export.py --include`         | Model
+---                         | ---                           | ---
+PyTorch                     | -                             | yolov5s.pt
+TorchScript                 | `torchscript`                 | yolov5s.torchscript
+ONNX                        | `onnx`                        | yolov5s.onnx
+OpenVINO                    | `openvino`                    | yolov5s_openvino_model/
+TensorRT                    | `engine`                      | yolov5s.engine
+CoreML                      | `coreml`                      | yolov5s.mlmodel
+TensorFlow SavedModel       | `saved_model`                 | yolov5s_saved_model/
+TensorFlow GraphDef         | `pb`                          | yolov5s.pb
+TensorFlow Lite             | `tflite`                      | yolov5s.tflite
+TensorFlow Edge TPU         | `edgetpu`                     | yolov5s_edgetpu.tflite
+TensorFlow.js               | `tfjs`                        | yolov5s_web_model/
+
+Requirements:
+    $ pip install -r requirements.txt coremltools onnx onnx-simplifier onnxruntime openvino-dev tensorflow-cpu  # CPU
+    $ pip install -r requirements.txt coremltools onnx onnx-simplifier onnxruntime-gpu openvino-dev tensorflow  # GPU
+
+Usage:
+    $ python utils/benchmarks.py --weights yolov5s.pt --img 640
+"""
+
+import argparse
+import sys
+import time
+from pathlib import Path
+
+import pandas as pd
+
+FILE = Path(__file__).resolve()
+ROOT = FILE.parents[1]  # YOLOv5 root directory
+if str(ROOT) not in sys.path:
+    sys.path.append(str(ROOT))  # add ROOT to PATH
+# ROOT = ROOT.relative_to(Path.cwd())  # relative
+
+import export
+import val
+from utils import notebook_init
+from utils.general import LOGGER, print_args
+
+
+def run(weights=ROOT / 'yolov5s.pt',  # weights path
+        imgsz=640,  # inference size (pixels)
+        batch_size=1,  # batch size
+        data=ROOT / 'data/coco128.yaml',  # dataset.yaml path
+        ):
+    y, t = [], time.time()
+    formats = export.export_formats()
+    for i, (name, f, suffix) in formats.iterrows():  # index, (name, file, suffix)
+        try:
+            w = weights if f == '-' else export.run(weights=weights, imgsz=[imgsz], include=[f], device='cpu')[-1]
+            assert suffix in str(w), 'export failed'
+            result = val.run(data, w, batch_size, imgsz=imgsz, plots=False, device='cpu', task='benchmark')
+            metrics = result[0]  # metrics (mp, mr, map50, map, *losses(box, obj, cls))
+            speeds = result[2]  # times (preprocess, inference, postprocess)
+            y.append([name, metrics[3], speeds[1]])  # mAP, t_inference
+        except Exception as e:
+            LOGGER.warning(f'WARNING: Benchmark failure for {name}: {e}')
+            y.append([name, None, None])  # mAP, t_inference
+
+    # Print results
+    LOGGER.info('\n')
+    parse_opt()
+    notebook_init()  # print system info
+    py = pd.DataFrame(y, columns=['Format', 'mAP@0.5:0.95', 'Inference time (ms)'])
+    LOGGER.info(f'\nBenchmarks complete ({time.time() - t:.2f}s)')
+    LOGGER.info(str(py))
+    return py
+
+
+def parse_opt():
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--weights', type=str, default=ROOT / 'yolov5s.pt', help='weights path')
+    parser.add_argument('--imgsz', '--img', '--img-size', type=int, default=640, help='inference size (pixels)')
+    parser.add_argument('--batch-size', type=int, default=1, help='batch size')
+    parser.add_argument('--data', type=str, default=ROOT / 'data/coco128.yaml', help='dataset.yaml path')
+    opt = parser.parse_args()
+    print_args(FILE.stem, opt)
+    return opt
+
+
+def main(opt):
+    run(**vars(opt))
+
+
+if __name__ == "__main__":
+    opt = parse_opt()
+    main(opt)

utils/callbacks.py

@@ -0,0 +1,79 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+"""
+Callback utils
+"""
+
+
+class Callbacks:
+    """"
+    Handles all registered callbacks for YOLOv5 Hooks
+    """
+
+    def __init__(self):
+        # Define the available callbacks
+        self._callbacks = {
+            'on_pretrain_routine_start': [],
+            'on_pretrain_routine_end': [],
+
+            'on_train_start': [],
+            'on_train_epoch_start': [],
+            'on_train_batch_start': [],
+            'optimizer_step': [],
+            'on_before_zero_grad': [],
+            'on_train_batch_end': [],
+            'on_train_epoch_end': [],
+
+            'on_val_start': [],
+            'on_val_batch_start': [],
+            'on_val_image_end': [],
+            'on_val_batch_end': [],
+            'on_val_end': [],
+
+            'on_fit_epoch_end': [],  # fit = train + val
+            'on_fit_epoch_end_prune': [],  # fit = train + val
+            'on_model_save': [],
+            'on_train_end': [],
+            'on_params_update': [],
+            'teardown': [],
+        }
+        self.stop_training = False  # set True to interrupt training
+
+    def register_action(self, hook, name='', callback=None):
+        """
+        Register a new action to a callback hook
+
+        Args:
+            hook        The callback hook name to register the action to
+            name        The name of the action for later reference
+            callback    The callback to fire
+        """
+        assert hook in self._callbacks, f"hook '{hook}' not found in callbacks {self._callbacks}"
+        assert callable(callback), f"callback '{callback}' is not callable"
+        self._callbacks[hook].append({'name': name, 'callback': callback})
+
+    def get_registered_actions(self, hook=None):
+        """"
+        Returns all the registered actions by callback hook
+
+        Args:
+            hook The name of the hook to check, defaults to all
+        """
+        if hook:
+            return self._callbacks[hook]
+        else:
+            return self._callbacks
+
+    def run(self, hook, *args, **kwargs):
+        """
+        Loop through the registered actions and fire all callbacks
+
+        Args:
+            hook The name of the hook to check, defaults to all
+            args Arguments to receive from YOLOv5
+            kwargs Keyword Arguments to receive from YOLOv5
+        """
+
+        assert hook in self._callbacks, f"hook '{hook}' not found in callbacks {self._callbacks}"
+
+        for logger in self._callbacks[hook]:
+            logger['callback'](*args, **kwargs)

utils/downloads.py

@@ -0,0 +1,153 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+"""
+Download utils
+"""
+
+import os
+import platform
+import subprocess
+import time
+import urllib
+from pathlib import Path
+from zipfile import ZipFile
+
+import requests
+import torch
+
+
+def gsutil_getsize(url=''):
+    # gs://bucket/file size https://cloud.google.com/storage/docs/gsutil/commands/du
+    s = subprocess.check_output(f'gsutil du {url}', shell=True).decode('utf-8')
+    return eval(s.split(' ')[0]) if len(s) else 0  # bytes
+
+
+def safe_download(file, url, url2=None, min_bytes=1E0, error_msg=''):
+    # Attempts to download file from url or url2, checks and removes incomplete downloads < min_bytes
+    file = Path(file)
+    assert_msg = f"Downloaded file '{file}' does not exist or size is < min_bytes={min_bytes}"
+    try:  # url1
+        print(f'Downloading {url} to {file}...')
+        torch.hub.download_url_to_file(url, str(file))
+        assert file.exists() and file.stat().st_size > min_bytes, assert_msg  # check
+    except Exception as e:  # url2
+        file.unlink(missing_ok=True)  # remove partial downloads
+        print(f'ERROR: {e}\nRe-attempting {url2 or url} to {file}...')
+        os.system(f"curl -L '{url2 or url}' -o '{file}' --retry 3 -C -")  # curl download, retry and resume on fail
+    finally:
+        if not file.exists() or file.stat().st_size < min_bytes:  # check
+            file.unlink(missing_ok=True)  # remove partial downloads
+            print(f"ERROR: {assert_msg}\n{error_msg}")
+        print('')
+
+
+def attempt_download(file, repo='ultralytics/yolov5'):  # from utils.downloads import *; attempt_download()
+    # Attempt file download if does not exist
+    file = Path(str(file).strip().replace("'", ''))
+
+    if not file.exists():
+        # URL specified
+        name = Path(urllib.parse.unquote(str(file))).name  # decode '%2F' to '/' etc.
+        if str(file).startswith(('http:/', 'https:/')):  # download
+            url = str(file).replace(':/', '://')  # Pathlib turns :// -> :/
+            file = name.split('?')[0]  # parse authentication https://url.com/file.txt?auth...
+            if Path(file).is_file():
+                print(f'Found {url} locally at {file}')  # file already exists
+            else:
+                safe_download(file=file, url=url, min_bytes=1E5)
+            return file
+
+        # GitHub assets
+        file.parent.mkdir(parents=True, exist_ok=True)  # make parent dir (if required)
+        try:
+            response = requests.get(f'https://api.github.com/repos/{repo}/releases/latest').json()  # github api
+            assets = [x['name'] for x in response['assets']]  # release assets, i.e. ['yolov5s.pt', 'yolov5m.pt', ...]
+            tag = response['tag_name']  # i.e. 'v1.0'
+        except Exception:  # fallback plan
+            assets = ['yolov5n.pt', 'yolov5s.pt', 'yolov5m.pt', 'yolov5l.pt', 'yolov5x.pt',
+                      'yolov5n6.pt', 'yolov5s6.pt', 'yolov5m6.pt', 'yolov5l6.pt', 'yolov5x6.pt']
+            try:
+                tag = subprocess.check_output('git tag', shell=True, stderr=subprocess.STDOUT).decode().split()[-1]
+            except Exception:
+                tag = 'v6.0'  # current release
+
+        if name in assets:
+            safe_download(file,
+                          url=f'https://github.com/{repo}/releases/download/{tag}/{name}',
+                          # url2=f'https://storage.googleapis.com/{repo}/ckpt/{name}',  # backup url (optional)
+                          min_bytes=1E5,
+                          error_msg=f'{file} missing, try downloading from https://github.com/{repo}/releases/')
+
+    return str(file)
+
+
+def gdrive_download(id='16TiPfZj7htmTyhntwcZyEEAejOUxuT6m', file='tmp.zip'):
+    # Downloads a file from Google Drive. from yolov5.utils.downloads import *; gdrive_download()
+    t = time.time()
+    file = Path(file)
+    cookie = Path('cookie')  # gdrive cookie
+    print(f'Downloading https://drive.google.com/uc?export=download&id={id} as {file}... ', end='')
+    file.unlink(missing_ok=True)  # remove existing file
+    cookie.unlink(missing_ok=True)  # remove existing cookie
+
+    # Attempt file download
+    out = "NUL" if platform.system() == "Windows" else "/dev/null"
+    os.system(f'curl -c ./cookie -s -L "drive.google.com/uc?export=download&id={id}" > {out}')
+    if os.path.exists('cookie'):  # large file
+        s = f'curl -Lb ./cookie "drive.google.com/uc?export=download&confirm={get_token()}&id={id}" -o {file}'
+    else:  # small file
+        s = f'curl -s -L -o {file} "drive.google.com/uc?export=download&id={id}"'
+    r = os.system(s)  # execute, capture return
+    cookie.unlink(missing_ok=True)  # remove existing cookie
+
+    # Error check
+    if r != 0:
+        file.unlink(missing_ok=True)  # remove partial
+        print('Download error ')  # raise Exception('Download error')
+        return r
+
+    # Unzip if archive
+    if file.suffix == '.zip':
+        print('unzipping... ', end='')
+        ZipFile(file).extractall(path=file.parent)  # unzip
+        file.unlink()  # remove zip
+
+    print(f'Done ({time.time() - t:.1f}s)')
+    return r
+
+
+def get_token(cookie="./cookie"):
+    with open(cookie) as f:
+        for line in f:
+            if "download" in line:
+                return line.split()[-1]
+    return ""
+
+# Google utils: https://cloud.google.com/storage/docs/reference/libraries ----------------------------------------------
+#
+#
+# def upload_blob(bucket_name, source_file_name, destination_blob_name):
+#     # Uploads a file to a bucket
+#     # https://cloud.google.com/storage/docs/uploading-objects#storage-upload-object-python
+#
+#     storage_client = storage.Client()
+#     bucket = storage_client.get_bucket(bucket_name)
+#     blob = bucket.blob(destination_blob_name)
+#
+#     blob.upload_from_filename(source_file_name)
+#
+#     print('File {} uploaded to {}.'.format(
+#         source_file_name,
+#         destination_blob_name))
+#
+#
+# def download_blob(bucket_name, source_blob_name, destination_file_name):
+#     # Uploads a blob from a bucket
+#     storage_client = storage.Client()
+#     bucket = storage_client.get_bucket(bucket_name)
+#     blob = bucket.blob(source_blob_name)
+#
+#     blob.download_to_filename(destination_file_name)
+#
+#     print('Blob {} downloaded to {}.'.format(
+#         source_blob_name,
+#         destination_file_name))