OpenPoseを試してみた。

OpenPose-Tensorflow

本家CaffeベースのOpenPoseを試したが・・・
Check failed: error == cudaSuccess (2 vs. 0) out of memoryで断念。
Tensorflow版はうまく動いた。

• 20/05/08
• Ubuntu18.04.4
• GeForce RTX 2060
• Docker version 19.03.8

REF

https://github.com/ildoonet/tf-pose-estimation

• dockerコンテナ上に環境を構築して実行した。

1. docker container

dockerfile

FROM nvidia/cuda:10.2-cudnn7-devel-ubuntu18.04

ENV DEBIAN_FRONTEND=noninteractive \
    LC_ALL=C.UTF-8 \
    LANG=C.UTF-8

RUN apt update && apt install -y --no-install-recommends \
    git curl wget \
    python3-dev \
    cython3 \
    python3-tk \
    libgtk2.0-dev \
    swig \
    imagemagick \
 && rm -rf /var/lib/apt/lists/*

RUN curl https://bootstrap.pypa.io/get-pip.py -o get-pip.py \
 && python3 get-pip.py \
 && rm get-pip.py

RUN pip3 install --no-cache-dir \
    numpy \
    tensorflow-gpu==1.15 \
    opencv-python==3.4.5.20

WORKDIR /workspace
RUN git clone https://www.github.com/ildoonet/tf-pose-estimation \
 && cd tf-pose-estimation \
 && pip3 install -r requirements.txt

WORKDIR /workspace/tf-pose-estimation/tf_pose/pafprocess
RUN swig -python -c++ pafprocess.i && python3 setup.py build_ext --inplace

WORKDIR /workspace/tf-pose-estimation/models/graph/cmu
RUN bash download.sh

build image

mkdir docker-build
cd docker-build

# dockerfileを作成
sudo gedit dockerfile

docker build -t openpose-tensorflow:1.15 .

run container

xhost +
docker run -it --rm --gpus all \
            -e DISPLAY=$DISPLAY \
            -v /tmp/.X11-unix:/tmp/.X11-unix \
            -v /etc/group:/etc/group:ro \
            -v /etc/passwd:/etc/passwd:ro \
            -u $(id -u $USER):$(id -g $USER) \
            -w /workspace/tf-pose-estimation/ \
            openpose-tensorflow:1.15

2. Run demo

python3 run.py --model=mobilenet_thin --resize=432x368 --image=./images/p1.jpg

result image

• INFO inference image: ./images/p1.jpg in 0.2125 seconds

result

以上。