Code Monkey home page Code Monkey logo

mmaction2-onnx-export-env's Introduction

mmaction2-onnx-export-env

1. Usage

1-1. Docker Pull

$ docker pull ghcr.io/pinto0309/mmaction2onnxexpenv:latest
$ git clone https://github.com/open-mmlab/mmaction2.git && cd mmaction2
$ git checkout 6ea00d23d66fd491bbf19317b4677bf6f396d26f
$ mkdir -p data
$ docker run --rm -it --gpus all --shm-size 10gb \
-v `pwd`:/home/user/workdir \
ghcr.io/pinto0309/mmaction2onnxexpenv:latest

1-2. Docker Build

$ git clone https://github.com/PINTO0309/mmaction2-onnx-export-env.git && cd mmaction2-onnx-export-env
$ git clone https://github.com/open-mmlab/mmaction2.git && cd mmaction2
$ git checkout 6ea00d23d66fd491bbf19317b4677bf6f396d26f
$ mkdir -p data
$ mv docker/Dockerfile docker/Dockerfile_org
$ cp ../Dockerfile docker/Dockerfile
$ docker build -t mmaction2onnxexpenv -f docker/Dockerfile .
$ docker run --rm -it --gpus all --shm-size 10gb \
-v `pwd`:/home/user/workdir \
ghcr.io/pinto0309/mmaction2onnxexpenv:latest

2. ONNX export syntax

- Recognizers
 - TSN: $batch $clip $channel $height $width  1 1 3 224 224
 - I3D: $batch $clip $channel $time $height $width  1 1 3 32 224 224

$ python tools/deployment/pytorch2onnx.py \
    ${CONFIG_FILE} \
    ${CHECKPOINT_FILE} \
    --shape ${SHAPE} \
    --verify \
    --output-file ${OUTPUT_FILE} \
    --opset-version 11

========================================
- Localizers
 - BSN etc: Default  1 1 3 224 224

$ python tools/deployment/pytorch2onnx.py \
    ${CONFIG_FILE} \
    ${CHECKPOINT_FILE} \
    --shape ${SHAPE} \
    --verify \
    --output-file ${OUTPUT_FILE} \
    --is-localizer \
    --opset-version 11

3. ONNX export sample

# PoseC3D - FineGYM/NTU60/NTU120
## config
wget https://github.com/open-mmlab/mmaction2/raw/master/configs/skeleton/posec3d/slowonly_r50_u48_240e_gym_keypoint.py
wget https://github.com/open-mmlab/mmaction2/raw/master/configs/skeleton/posec3d/slowonly_r50_u48_240e_gym_limb.py
wget https://github.com/open-mmlab/mmaction2/raw/master/configs/skeleton/posec3d/slowonly_r50_u48_240e_ntu60_xsub_keypoint.py
wget https://github.com/open-mmlab/mmaction2/raw/master/configs/skeleton/posec3d/slowonly_r50_u48_240e_ntu60_xsub_limb.py
wget https://github.com/open-mmlab/mmaction2/raw/master/configs/skeleton/posec3d/slowonly_r50_u48_240e_ntu120_xsub_keypoint.py
wget https://github.com/open-mmlab/mmaction2/raw/master/configs/skeleton/posec3d/slowonly_r50_u48_240e_ntu120_xsub_limb.py

## checkpoint
wget https://download.openmmlab.com/mmaction/skeleton/posec3d/slowonly_r50_u48_240e_gym_keypoint/slowonly_r50_u48_240e_gym_keypoint-b07a98a0.pth -O slowonly_r50_u48_240e_gym_keypoint.pth
wget https://download.openmmlab.com/mmaction/skeleton/posec3d/slowonly_r50_u48_240e_gym_limb/slowonly_r50_u48_240e_gym_limb-c0d7b482.pth -O slowonly_r50_u48_240e_gym_limb.pth
wget https://download.openmmlab.com/mmaction/skeleton/posec3d/slowonly_r50_u48_240e_ntu60_xsub_keypoint/slowonly_r50_u48_240e_ntu60_xsub_keypoint-f3adabf1.pth -O slowonly_r50_u48_240e_ntu60_xsub_keypoint.pth
wget https://download.openmmlab.com/mmaction/skeleton/posec3d/slowonly_r50_u48_240e_ntu60_xsub_limb/slowonly_r50_u48_240e_ntu60_xsub_limb-1d69006a.pth -O slowonly_r50_u48_240e_ntu60_xsub_limb.pth
wget https://download.openmmlab.com/mmaction/skeleton/posec3d/slowonly_r50_u48_240e_ntu120_xsub_keypoint/slowonly_r50_u48_240e_ntu120_xsub_keypoint-6736b03f.pth -O slowonly_r50_u48_240e_ntu120_xsub_keypoint.pth
wget https://download.openmmlab.com/mmaction/skeleton/posec3d/slowonly_r50_u48_240e_ntu120_xsub_limb/slowonly_r50_u48_240e_ntu120_xsub_limb-803c2317.pth -O slowonly_r50_u48_240e_ntu120_xsub_limb.pth

KEYPOINTS=17 \
&& FRAMES=30 \
&& H=64 \
&& W=64 \
&& BATCH=1 \
&& CLIP=2

MODEL=slowonly_r50_u48_240e_gym_keypoint \
&& python tools/deployment/pytorch2onnx.py \
${MODEL}.py \
${MODEL}.pth \
--shape ${BATCH} ${CLIP} ${KEYPOINTS} ${FRAMES} ${H} ${W} \
--verify \
--output-file ${MODEL}_${BATCH}x${CLIP}x${KEYPOINTS}x${FRAMES}x${H}x${W}.onnx \
--opset-version 11 \
--softmax \
&& onnxsim ${MODEL}_${BATCH}x${CLIP}x${KEYPOINTS}x${FRAMES}x${H}x${W}.onnx ${MODEL}_${BATCH}x${CLIP}x${KEYPOINTS}x${FRAMES}x${H}x${W}.onnx

MODEL=slowonly_r50_u48_240e_gym_limb \
&& python tools/deployment/pytorch2onnx.py \
${MODEL}.py \
${MODEL}.pth \
--shape ${BATCH} ${CLIP} ${KEYPOINTS} ${FRAMES} ${H} ${W} \
--verify \
--output-file ${MODEL}_${BATCH}x${CLIP}x${KEYPOINTS}x${FRAMES}x${H}x${W}.onnx \
--opset-version 11 \
--softmax \
&& onnxsim ${MODEL}_${BATCH}x${CLIP}x${KEYPOINTS}x${FRAMES}x${H}x${W}.onnx ${MODEL}_${BATCH}x${CLIP}x${KEYPOINTS}x${FRAMES}x${H}x${W}.onnx


MODEL=slowonly_r50_u48_240e_ntu60_xsub_keypoint \
&& python tools/deployment/pytorch2onnx.py \
${MODEL}.py \
${MODEL}.pth \
--shape ${BATCH} ${CLIP} ${KEYPOINTS} ${FRAMES} ${H} ${W} \
--verify \
--output-file ${MODEL}_${BATCH}x${CLIP}x${KEYPOINTS}x${FRAMES}x${H}x${W}.onnx \
--opset-version 11 \
--softmax \
&& onnxsim ${MODEL}_${BATCH}x${CLIP}x${KEYPOINTS}x${FRAMES}x${H}x${W}.onnx ${MODEL}_${BATCH}x${CLIP}x${KEYPOINTS}x${FRAMES}x${H}x${W}.onnx

MODEL=slowonly_r50_u48_240e_ntu60_xsub_limb \
&& python tools/deployment/pytorch2onnx.py \
${MODEL}.py \
${MODEL}.pth \
--shape ${BATCH} ${CLIP} ${KEYPOINTS} ${FRAMES} ${H} ${W} \
--verify \
--output-file ${MODEL}_${BATCH}x${CLIP}x${KEYPOINTS}x${FRAMES}x${H}x${W}.onnx \
--opset-version 11 \
--softmax \
&& onnxsim ${MODEL}_${BATCH}x${CLIP}x${KEYPOINTS}x${FRAMES}x${H}x${W}.onnx ${MODEL}_${BATCH}x${CLIP}x${KEYPOINTS}x${FRAMES}x${H}x${W}.onnx


MODEL=slowonly_r50_u48_240e_ntu120_xsub_keypoint \
&& python tools/deployment/pytorch2onnx.py \
${MODEL}.py \
${MODEL}.pth \
--shape ${BATCH} ${CLIP} ${KEYPOINTS} ${FRAMES} ${H} ${W} \
--verify \
--output-file ${MODEL}_${BATCH}x${CLIP}x${KEYPOINTS}x${FRAMES}x${H}x${W}.onnx \
--opset-version 11 \
--softmax \
&& onnxsim ${MODEL}_${BATCH}x${CLIP}x${KEYPOINTS}x${FRAMES}x${H}x${W}.onnx ${MODEL}_${BATCH}x${CLIP}x${KEYPOINTS}x${FRAMES}x${H}x${W}.onnx

MODEL=slowonly_r50_u48_240e_ntu120_xsub_limb \
&& python tools/deployment/pytorch2onnx.py \
${MODEL}.py \
${MODEL}.pth \
--shape ${BATCH} ${CLIP} ${KEYPOINTS} ${FRAMES} ${H} ${W} \
--verify \
--output-file ${MODEL}_${BATCH}x${CLIP}x${KEYPOINTS}x${FRAMES}x${H}x${W}.onnx \
--opset-version 11 \
--softmax \
&& onnxsim ${MODEL}_${BATCH}x${CLIP}x${KEYPOINTS}x${FRAMES}x${H}x${W}.onnx ${MODEL}_${BATCH}x${CLIP}x${KEYPOINTS}x${FRAMES}x${H}x${W}.onnx

4. Model Type - ModelZoo OVERVIEW

  1. Spatio Temporal Action Detection Models
  2. Action Localization Models
  3. Action Recognition Models
  4. Skeleton-based Action Recognition Models

https://mmaction2.readthedocs.io/en/latest/modelzoo.html

5. MMAction2

https://github.com/open-mmlab/mmaction2

mmaction2-onnx-export-env's People

Contributors

pinto0309 avatar

Stargazers

avatar avatar avatar avatar avatar avatar avatar

Watchers

avatar avatar avatar

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.