Release - Dockerfiles with Fixed Model Downloads - CPU / CPU + CUDA + cuDNN

[hiibolt]

Issue Summary

It seems that something behind the scenes at CMU broke, as their CDN service which a manual build of OpenPose requires to obtain models is currently down.

As a result, OpenPose is currently unusable. However, thanks to the work done in Issue #1567, it's possible to instead disable the model downloads and instead use a third party hosting of the models on DropBox.

CMake also seems to have issues with building to support CMake, which the work here seems to fix - however, it targets a now depreciated version of nvidia/cuda. By changing the target from nvidia/cuda:11.4.0-cudnn8-devel-ubuntu18.04 to nvidia/cuda:11.3.1-cudnn8-devel-ubuntu18.04, it now has a functional source image.

By combining these two fixes, you can create both a CPU + Python API and a CPU + Python API + CUDA + cuDNN Docker image.

CPU + Python API: (Link)

# https://hub.docker.com/r/cwaffles/openpose
FROM ubuntu:focal

#get deps
RUN apt-get update && \
    DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends \
    python3-dev python3-pip python3-setuptools git g++ wget make libprotobuf-dev protobuf-compiler libopencv-dev \
    libgoogle-glog-dev libboost-all-dev caffe-cpu libhdf5-dev libatlas-base-dev

#for python api
RUN pip3 install --upgrade pip
RUN pip3 install numpy opencv-python 

#replace cmake as old version has CUDA variable bugs
RUN wget https://github.com/Kitware/CMake/releases/download/v3.16.0/cmake-3.16.0-Linux-x86_64.tar.gz && \
tar xzf cmake-3.16.0-Linux-x86_64.tar.gz -C /opt && \
rm cmake-3.16.0-Linux-x86_64.tar.gz
ENV PATH="/opt/cmake-3.16.0-Linux-x86_64/bin:${PATH}"

#get openpose
RUN git clone https://github.com/KWNahyun/openpose.git
WORKDIR /openpose
RUN    cd /openpose/models/pose/body_25 && wget -O pose_iter_584000.caffemodel -c https://www.dropbox.com/s/3x0xambj2rkyrap/pose_iter_584000.caffemodel?dl=0
# pose_iter_116000.caffemodel
RUN    cd /openpose/models/face && wget -O pose_iter_116000.caffemodel-c https://www.dropbox.com/s/d08srojpvwnk252/pose_iter_116000.caffemodel?dl=0
# pose_iter_102000.caffemodel
RUN    cd /openpose/models/hand && wget -O pose_iter_102000.caffemodel -c https://www.dropbox.com/s/gqgsme6sgoo0zxf/pose_iter_102000.caffemodel?dl=0

# use 'sed' to comment out the line in the OpenPose repo that downloads the model from the failed link
RUN sed -i 's/executeShInItsFolder "getModels.sh"/# executeShInItsFolder "getModels.sh"/g' /openpose/scripts/ubuntu/install_openpose_JetsonTX2_JetPack3.1.sh
RUN sed -i 's/executeShInItsFolder "getModels.sh"/# executeShInItsFolder "getModels.sh"/g' /openpose/scripts/ubuntu/install_openpose_JetsonTX2_JetPack3.3.sh
RUN sed -i 's/download_model("BODY_25"/# download_model("BODY_25"/g' /openpose/CMakeLists.txt
RUN sed -i 's/78287B57CF85FA89C03F1393D368E5B7/# 78287B57CF85FA89C03F1393D368E5B7/g' /openpose/CMakeLists.txt
RUN sed -i 's/download_model("body (COCO)"/# download_model("body (COCO)"/g' /openpose/CMakeLists.txt
RUN sed -i 's/5156d31f670511fce9b4e28b403f2939/# 5156d31f670511fce9b4e28b403f2939/g' /openpose/CMakeLists.txt
RUN sed -i 's/download_model("body (MPI)"/# download_model("body (MPI)"/g' /openpose/CMakeLists.txt
RUN sed -i 's/2ca0990c7562bd7ae03f3f54afa96e00/# 2ca0990c7562bd7ae03f3f54afa96e00/g' /openpose/CMakeLists.txt
RUN sed -i 's/download_model("face"/# download_model("face"/g' /openpose/CMakeLists.txt
RUN sed -i 's/e747180d728fa4e4418c465828384333/# e747180d728fa4e4418c465828384333/g' /openpose/CMakeLists.txt
RUN sed -i 's/download_model("hand"/# download_model("hand"/g' /openpose/CMakeLists.txt
RUN sed -i 's/a82cfc3fea7c62f159e11bd3674c1531/# a82cfc3fea7c62f159e11bd3674c1531/g' /openpose/CMakeLists.txt

#build it
WORKDIR /openpose/build
RUN cmake -DBUILD_PYTHON=ON -DGPU_MODE=CPU_ONLY -DOWNLOAD_HAND_MODEL=OFF -DOWNLOAD_FACE_MODEL=OFF .. && make -j `nproc`

CPU + Python API + CUDA + cuDNN (Link)

# https://hub.docker.com/r/cwaffles/openpose
FROM nvidia/cuda:11.3.1-cudnn8-devel-ubuntu18.04

#get deps
RUN apt-get update && \
DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends \
python3-dev python3-pip python3-setuptools git g++ wget make libprotobuf-dev protobuf-compiler libopencv-dev \
libgoogle-glog-dev libboost-all-dev libcaffe-cuda-dev libhdf5-dev libatlas-base-dev

#for python api
RUN pip3 install --upgrade pip
RUN pip3 install numpy opencv-python 

#replace cmake as old version has CUDA variable bugs
RUN wget https://github.com/Kitware/CMake/releases/download/v3.16.0/cmake-3.16.0-Linux-x86_64.tar.gz && \
tar xzf cmake-3.16.0-Linux-x86_64.tar.gz -C /opt && \
rm cmake-3.16.0-Linux-x86_64.tar.gz
ENV PATH="/opt/cmake-3.16.0-Linux-x86_64/bin:${PATH}"

#get openpose
WORKDIR /openpose
RUN git clone https://github.com/CMU-Perceptual-Computing-Lab/openpose.git .

RUN    cd /openpose/models/pose/body_25 && wget -O pose_iter_584000.caffemodel -c https://www.dropbox.com/s/3x0xambj2rkyrap/pose_iter_584000.caffemodel?dl=0
# pose_iter_116000.caffemodel
RUN    cd /openpose/models/face && wget -O pose_iter_116000.caffemodel-c https://www.dropbox.com/s/d08srojpvwnk252/pose_iter_116000.caffemodel?dl=0
# pose_iter_102000.caffemodel
RUN    cd /openpose/models/hand && wget -O pose_iter_102000.caffemodel -c https://www.dropbox.com/s/gqgsme6sgoo0zxf/pose_iter_102000.caffemodel?dl=0

# use 'sed' to comment out the line in the OpenPose repo that downloads the model from the failed link
RUN sed -i 's/executeShInItsFolder "getModels.sh"/# executeShInItsFolder "getModels.sh"/g' /openpose/scripts/ubuntu/install_openpose_JetsonTX2_JetPack3.1.sh
RUN sed -i 's/executeShInItsFolder "getModels.sh"/# executeShInItsFolder "getModels.sh"/g' /openpose/scripts/ubuntu/install_openpose_JetsonTX2_JetPack3.3.sh
RUN sed -i 's/download_model("BODY_25"/# download_model("BODY_25"/g' /openpose/CMakeLists.txt
RUN sed -i 's/78287B57CF85FA89C03F1393D368E5B7/# 78287B57CF85FA89C03F1393D368E5B7/g' /openpose/CMakeLists.txt
RUN sed -i 's/download_model("body (COCO)"/# download_model("body (COCO)"/g' /openpose/CMakeLists.txt
RUN sed -i 's/5156d31f670511fce9b4e28b403f2939/# 5156d31f670511fce9b4e28b403f2939/g' /openpose/CMakeLists.txt
RUN sed -i 's/download_model("body (MPI)"/# download_model("body (MPI)"/g' /openpose/CMakeLists.txt
RUN sed -i 's/2ca0990c7562bd7ae03f3f54afa96e00/# 2ca0990c7562bd7ae03f3f54afa96e00/g' /openpose/CMakeLists.txt
RUN sed -i 's/download_model("face"/# download_model("face"/g' /openpose/CMakeLists.txt
RUN sed -i 's/e747180d728fa4e4418c465828384333/# e747180d728fa4e4418c465828384333/g' /openpose/CMakeLists.txt
RUN sed -i 's/download_model("hand"/# download_model("hand"/g' /openpose/CMakeLists.txt
RUN sed -i 's/a82cfc3fea7c62f159e11bd3674c1531/# a82cfc3fea7c62f159e11bd3674c1531/g' /openpose/CMakeLists.txt

#build it
WORKDIR /openpose/build
RUN cmake -DBUILD_PYTHON=ON .. && make -j `nproc`
WORKDIR /openpose

Type of Issue

• Enhancement / offering possible extensions / pull request / etc

[hiibolt]

Dockerfiles, example usage, and expected output can be found here:

• https://github.com/hiibolt/igait-openpose/blob/main/README.md

CPU + Python API + CUDA + cuDNN Docker Image:

• https://github.com/users/hiibolt/packages/container/package/igait-openpose
• ghcr.io/hiibolt/igait-openpose:latest

[xiangyann]

Nice! But I think the problem that needs to be fixed first is the performance degradation using cudnn>=8, or else a docker update might not be any useful.
Does 18.04+CUDA11+cudnn8 make a performant combo?

[hiibolt]

Nice! But I think the problem that needs to be fixed first is the performance degradation using cudnn>=8, or else a docker update might not be any useful. Does 18.04+CUDA11+cudnn8 make a performant combo?

It has been very performant for our use case so far, the computation time for x2 57 second videos at 1080p (~60MB each) is less than a minute. Sadly, I do not have computation times for CuDNN<8 to compare to.

Specs:
Allocated CPU: 1 VCPU from AMD EPYC 7713
Allocated RAM: 1GB from 512GB pool
Allocated GPU: NVIDIA A100
Launch Arguments: --video "..." --display 0 --write_video "..." --write_json "..."

[ :3 - Starting OpenPose pose estimation... - :3 ]
Starting OpenPose demo...
Configuring OpenPose...
Starting thread(s)...
Auto-detecting all available GPUs... Detected 1 GPU(s), using 1 of them starting at GPU 0.
Empty frame detected, frame number 0 of 479. In /openpose/src/openpose/producer/producer.cpp:checkFrameIntegrity():290
OpenCV: FFMPEG: tag 0x47504a4d/'MJPG' is not supported with codec id 8 and format 'mov / QuickTime / MOV'
OpenCV: FFMPEG: fallback to use tag 0x6765706a/'jpeg'
OpenPose demo successfully finished. Total time: 19.738495 seconds.
Starting OpenPose demo...
Configuring OpenPose...
Starting thread(s)...
Auto-detecting all available GPUs... Detected 1 GPU(s), using 1 of them starting at GPU 0.
OpenCV: FFMPEG: tag 0x47504a4d/'MJPG' is not supported with codec id 8 and format 'mov / QuickTime / MOV'
OpenCV: FFMPEG: fallback to use tag 0x6765706a/'jpeg'
OpenPose demo successfully finished. Total time: 25.475751 seconds.
[ :3 - Finished OpenPose pose estimations! - :3 ]