LoginSignup
0
1

More than 1 year has passed since last update.

@electronics_diy721(elec tron)

pytorch-openposeで姿勢推定（MacOS CPU）

Last updated at Posted at 2021-07-19

今回やったこと

スクリーンショット 2021-07-19 21.18.33.png

スクリーンショット 2021-07-19 21.36.06.png

Pytorch-openposeを利用

当初、次のリポジトリの資源をgit cloneしようとしたが、tesorflow=1.4.1が入れられなかったため、利用を諦めた。

変わりに、PyTorch実装の資源を次から入れることにした。

Terminal
electron@diynoMacBook-Pro Desktop % git clone https://github.com/Hzzone/pytorch-openpose.git
Cloning into 'pytorch-openpose'...
remote: Enumerating objects: 154, done.
remote: Counting objects: 100% (3/3), done.
remote: Compressing objects: 100% (3/3), done.
remote: Total 154 (delta 0), reused 0 (delta 0), pack-reused 151
Receiving objects: 100% (154/154), 20.18 MiB | 13.12 MiB/s, done.
Resolving deltas: 100% (68/68), done.
electron@diynoMacBook-Pro Desktop %
Terminal
electron@diynoMacBook-Pro Desktop % cd pytorch-openpose 
electron@diynoMacBook-Pro pytorch-openpose % ls
README.md       demo_camera.py      images          notebooks       src
demo.py         demo_video.py       model           requirements.txt
electron@diynoMacBook-Pro pytorch-openpose %

以下から、body_pose_model.pth hand_pose_model.pthをダウンロード

両ファイルを、./model直下に移動

Terminal
electron@diynoMacBook-Pro pytorch-openpose % ls model
electron@diynoMacBook-Pro pytorch-openpose % ls *.pth
body_pose_model.pth hand_pose_model.pth
electron@diynoMacBook-Pro pytorch-openpose % 
electron@diynoMacBook-Pro pytorch-openpose % mv *.pth ./model/
electron@diynoMacBook-Pro pytorch-openpose % ls *.pth         
zsh: no matches found: *.pth
electron@diynoMacBook-Pro pytorch-openpose %
Terminal
electron@diynoMacBook-Pro pytorch-openpose % ls model         
body_pose_model.pth hand_pose_model.pth
electron@diynoMacBook-Pro pytorch-openpose %
Terminal
electron@diynoMacBook-Pro pytorch-openpose % python3 demo_camera.py                                                           
Traceback (most recent call last):
  File "/Users/electron/Desktop/pytorch-openpose/demo_camera.py", line 5, in <module>
    import torch
  File "/usr/local/lib/python3.9/site-packages/torch/__init__.py", line 196, in <module>
    from torch._C import *
RuntimeError: module compiled against API version 0xe but this version of numpy is 0xd
electron@diynoMacBook-Pro pytorch-openpose %

以下のエラーメッセージが出た。

RuntimeError: module compiled against API version 0xe but this version of numpy is 0xd

Googleで調べると、numpyのバージョンに問題があるらしい。
requirement.txtに列挙されているnumpuのバージョン番号を見てみる。

Terminal
electron@diynoMacBook-Pro pytorch-openpose % cat requirements.txt 
numpy
matplotlib
opencv-python
scipy
scikit-image
tqdm%   
electron@diynoMacBook-Pro pytorch-openpose %

バージョンは特に指定されていない。

以下のサイトから、test.pyを写経。

test.py
import os
import cv2
import copy
from src import util
from src.body import Body

INPUT_FILE_NAME = "test.jpg"

if __name__ == "__main__":

    body_estimation = Body('model/body_pose_model.pth')

    target_image_path = 'images/' + INPUT_FILE_NAME
    oriImg = cv2.imread(target_image_path)  # B,G,R order
    candidate, subset = body_estimation(oriImg)
    canvas = copy.deepcopy(oriImg)
    canvas = util.draw_bodypose(canvas, candidate, subset)

    basename_name = os.path.splitext(os.path.basename(target_image_path))[0]

    result_image_path = "result/pose_" + basename_name + ".jpg"
    cv2.imwrite(result_image_path, canvas)

imageファイルをimages直下に格納。
ファイル名は、pose_test.jpg

スクリーンショット 2021-07-19 21.16.40.png

Terminal
electron@diynoMacBook-Pro pytorch-openpose % ls images | grep test 
test.jpg
electron@diynoMacBook-Pro pytorch-openpose %

やはり同じエラーが出る。

Terminal
electron@diynoMacBook-Pro pytorch-openpose % python3 test.py
Traceback (most recent call last):
  File "/Users/electron/Desktop/pytorch-openpose/test.py", line 5, in <module>
    from src.body import Body
  File "/Users/electron/Desktop/pytorch-openpose/src/body.py", line 8, in <module>
    import torch
  File "/usr/local/lib/python3.9/site-packages/torch/__init__.py", line 196, in <module>
    from torch._C import *
RuntimeError: module compiled against API version 0xe but this version of numpy is 0xd
electron@diynoMacBook-Pro pytorch-openpose %

Python 3.7.0に切り替える

Terminal
electron@diynoMacBook-Pro pytorch-openpose % pyenv local 3.7.0
electron@diynoMacBook-Pro pytorch-openpose % pyenv versions   
  system
  2.7.16
  3.6.0
  3.6.1
  3.6.3
  3.6.3/envs/gpt2_ja
  3.6.6
* 3.7.0 (set by /Users/electron/Desktop/pytorch-openpose/.python-version)
  3.7.4
  3.9.0
  gpt2_ja
electron@diynoMacBook-Pro pytorch-openpose %

改めて、依存ライブラリをinstall

Terminal
electron@diynoMacBook-Pro pytorch-openpose % pip3 install -r requirements.txt
Requirement already satisfied: numpy in /usr/local/lib/python3.9/site-packages (from -r requirements.txt (line 1)) (1.19.5)
Requirement already satisfied: matplotlib in /usr/local/lib/python3.9/site-packages (from -r requirements.txt (line 2)) (3.4.2)
Requirement already satisfied: opencv-python in /usr/local/lib/python3.9/site-packages (from -r requirements.txt (line 3)) (4.5.3.56)
Requirement already satisfied: scipy in /usr/local/lib/python3.9/site-packages (from -r requirements.txt (line 4)) (1.7.0)
Requirement already satisfied: scikit-image in /usr/local/lib/python3.9/site-packages (from -r requirements.txt (line 5)) (0.18.2)
Requirement already satisfied: tqdm in /usr/local/lib/python3.9/site-packages (from -r requirements.txt (line 6)) (4.61.2)
Requirement already satisfied: pillow>=6.2.0 in /usr/local/lib/python3.9/site-packages (from matplotlib->-r requirements.txt (line 2)) (8.3.1)
Requirement already satisfied: pyparsing>=2.2.1 in /usr/local/lib/python3.9/site-packages (from matplotlib->-r requirements.txt (line 2)) (2.4.7)
Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.9/site-packages (from matplotlib->-r requirements.txt (line 2)) (0.10.0)
Requirement already satisfied: kiwisolver>=1.0.1 in /usr/local/lib/python3.9/site-packages (from matplotlib->-r requirements.txt (line 2)) (1.3.1)
Requirement already satisfied: python-dateutil>=2.7 in /usr/local/lib/python3.9/site-packages (from matplotlib->-r requirements.txt (line 2)) (2.8.1)
Requirement already satisfied: six in /usr/local/lib/python3.9/site-packages (from cycler>=0.10->matplotlib->-r requirements.txt (line 2)) (1.15.0)
Requirement already satisfied: networkx>=2.0 in /usr/local/lib/python3.9/site-packages (from scikit-image->-r requirements.txt (line 5)) (2.5.1)
Requirement already satisfied: imageio>=2.3.0 in /usr/local/lib/python3.9/site-packages (from scikit-image->-r requirements.txt (line 5)) (2.9.0)
Requirement already satisfied: tifffile>=2019.7.26 in /usr/local/lib/python3.9/site-packages (from scikit-image->-r requirements.txt (line 5)) (2021.7.2)
Requirement already satisfied: PyWavelets>=1.1.1 in /usr/local/lib/python3.9/site-packages (from scikit-image->-r requirements.txt (line 5)) (1.1.1)
Requirement already satisfied: decorator<5,>=4.3 in /usr/local/lib/python3.9/site-packages (from networkx>=2.0->scikit-image->-r requirements.txt (line 5)) (4.4.2)
electron@diynoMacBook-Pro pytorch-openpose %

またダメ

Terminal
electron@diynoMacBook-Pro pytorch-openpose % python3 demo_camera.py
Traceback (most recent call last):
  File "/Users/electron/Desktop/pytorch-openpose/demo_camera.py", line 5, in <module>
    import torch
  File "/usr/local/lib/python3.9/site-packages/torch/__init__.py", line 196, in <module>
    from torch._C import *
RuntimeError: module compiled against API version 0xe but this version of numpy is 0xd
electron@diynoMacBook-Pro pytorch-openpose %

以下の次のアドバイスを採用

pip install numpy --upgrade works, something has to be updated in the files, when you figure out what, could you tell me please, thanks

Terminal
electron@diynoMacBook-Pro pytorch-openpose % pip3 install numpy --upgrade

（・・・省略・・・）

tensorflow 2.5.0 requires numpy~=1.19.2, but you have numpy 1.21.1 which is incompatible.
Successfully installed numpy-1.21.1
electron@diynoMacBook-Pro pytorch-openpose %

今度は、cudaがないエラー

Terminal
electron@diynoMacBook-Pro pytorch-openpose % python3 demo_camera.py
Traceback (most recent call last):
  File "/Users/electron/Desktop/pytorch-openpose/demo_camera.py", line 15, in <module>
    print(f"Torch device: {torch.cuda.get_device_name()}")
  File "/usr/local/lib/python3.9/site-packages/torch/cuda/__init__.py", line 276, in get_device_name
    return get_device_properties(device).name
  File "/usr/local/lib/python3.9/site-packages/torch/cuda/__init__.py", line 306, in get_device_properties
    _lazy_init()  # will define _get_device_properties
  File "/usr/local/lib/python3.9/site-packages/torch/cuda/__init__.py", line 164, in _lazy_init
    raise AssertionError("Torch not compiled with CUDA enabled")
AssertionError: Torch not compiled with CUDA enabled
electron@diynoMacBook-Pro pytorch-openpose %

以下のサイトの次のアドバイスを採用

AssertionError: Torch not compiled with CUDA enabled
はい、cudaね。test_video.pyの110行目と134行目に.cuda()があるので、.cuda()だけを削除し、再び実行。

以下をコメントアウト

print(f"Torch device: {torch.cuda.get_device_name()}")

Terminal
electron@diynoMacBook-Pro pytorch-openpose % cp demo_camera.py demo_camera_cpu.py

今度はちゃんと動いた:sunny:

Webカメラの画像が動画で表示され、自分の顔に骨格線が重畳された画像が表示された。

体を動かすと、ややタイムラグを伴って、骨格線も追従してついてきた。

CTRL-Zで強制終了

Terminal
electron@diynoMacBook-Pro pytorch-openpose % python3 demo_camera_cpu.py
^Z
zsh: suspended  python3 demo_camera_cpu.py
electron@diynoMacBook-Pro pytorch-openpose %

今度は、test.pyもうまく動いた

Terminal
electron@diynoMacBook-Pro pytorch-openpose % python3 test.py

resultディレクトリ直下に、解析後の画像ファイルが出力されている。

  • ファイル名：pose + 元のファイル名 * .jpg*
Terminal
electron@diynoMacBook-Pro pytorch-openpose % ls result
pose_test.jpg

electron@diynoMacBook-Pro pytorch-openpose %

スクリーンショット 2021-07-19 20.23.14.png

スクリーンショット 2021-07-19 20.23.22.png

demo.pyもうまく動いた

Terminal
electron@diynoMacBook-Pro pytorch-openpose % python3 demo.py
electron@diynoMacBook-Pro pytorch-openpose %

数秒後にウインドウが立ち上がり、次の画像が表示された。

スクリーンショット 2021-07-19 20.14.23.png

動画は実行に成功しなかった。

Terminal
electron@diynoMacBook-Pro pytorch-openpose % pip3 install ffmpeg
Terminal
electron@diynoMacBook-Pro pytorch-openpose % python3 demo_video.py dance.mp4                                
ffmpeg version 4.4 Copyright (c) 2000-2021 the FFmpeg developers
  built with Apple clang version 12.0.0 (clang-1200.0.32.29)
  configuration: --prefix=/usr/local/Cellar/ffmpeg/4.4_2 --enable-shared --enable-pthreads --enable-version3 --cc=clang --host-cflags= --host-ldflags= --enable-ffplay --enable-gnutls --enable-gpl --enable-libaom --enable-libbluray --enable-libdav1d --enable-libmp3lame --enable-libopus --enable-librav1e --enable-librubberband --enable-libsnappy --enable-libsrt --enable-libtesseract --enable-libtheora --enable-libvidstab --enable-libvorbis --enable-libvpx --enable-libwebp --enable-libx264 --enable-libx265 --enable-libxml2 --enable-libxvid --enable-lzma --enable-libfontconfig --enable-libfreetype --enable-frei0r --enable-libass --enable-libopencore-amrnb --enable-libopencore-amrwb --enable-libopenjpeg --enable-libspeex --enable-libsoxr --enable-libzmq --enable-libzimg --disable-libjack --disable-indev=jack --enable-avresample --enable-videotoolbox
  libavutil      56. 70.100 / 56. 70.100
  libavcodec     58.134.100 / 58.134.100
  libavformat    58. 76.100 / 58. 76.100
  libavdevice    58. 13.100 / 58. 13.100
  libavfilter     7.110.100 /  7.110.100
  libavresample   4.  0.  0 /  4.  0.  0
  libswscale      5.  9.100 /  5.  9.100
  libswresample   3.  9.100 /  3.  9.100
  libpostproc    55.  9.100 / 55.  9.100
Input #0, rawvideo, from 'pipe:':
  Duration: N/A, start: 0.000000, bitrate: 294796 kb/s
  Stream #0:0: Video: rawvideo (BGR[24] / 0x18524742), bgr24, 853x480, 294796 kb/s, 30 tbr, 30 tbn, 30 tbc
Stream mapping:
  Stream #0:0 -> #0:0 (rawvideo (native) -> h264 (libx264))
[libx264 @ 0x7f8c14809a00] width not divisible by 2 (853x480)
Error initializing output stream 0:0 -- Error while opening encoder for output stream #0:0 - maybe incorrect parameters such as bit_rate, rate, width or height
Conversion failed!
Traceback (most recent call last):
  File "/Users/electron/Desktop/pytorch-openpose/demo_video.py", line 132, in <module>
    writer(posed_frame)
  File "/Users/electron/Desktop/pytorch-openpose/demo_video.py", line 108, in __call__
    self.ff_proc.stdin.write(frame.tobytes())
BrokenPipeError: [Errno 32] Broken pipe
electron@diynoMacBook-Pro pytorch-openpose %

ウインドウが立ち上がり、次の１フレームで止まった状態で画像が表示された後、落ちた。

スクリーンショット 2021-07-19 21.10.52.png

test.pyを書き換えて、もう２つ、画像ファイルを加工してみる

取り扱う画像ファイルのパスを、コマンドライン引数で受け取るように書き換える。

test2.py
import os
import cv2
import copy
from src import util
from src.body import Body
import argparse    

# 動画ファイル名をコマンドライン引数から受け取る
parser = argparse.ArgumentParser(description='')    #
parser.add_argument('file_name')  
args = parser.parse_args()  

file = args.file_name


if __name__ == "__main__":

    body_estimation = Body('model/body_pose_model.pth')

    target_image_path = file
    oriImg = cv2.imread(target_image_path)  # B,G,R order
    candidate, subset = body_estimation(oriImg)
    canvas = copy.deepcopy(oriImg)
    canvas = util.draw_bodypose(canvas, candidate, subset)

    basename_name = os.path.splitext(os.path.basename(target_image_path))[0]

    result_image_path = "result/pose_" + basename_name + ".jpg"
    cv2.imwrite(result_image_path, canvas)
Terminal
electron@diynoMacBook-Pro pytorch-openpose % ls images/21stEmon.jpg                                            
images/21stEmon.jpg
electron@diynoMacBook-Pro pytorch-openpose %

スクリーンショット 2021-07-19 21.28.32.png

Terminal
electron@diynoMacBook-Pro pytorch-openpose %  python3 test2.py ./images/21stEmon.jpg                         
electron@diynoMacBook-Pro pytorch-openpose %  ls result/            
pose_21stEmon.jpg   pose_test.jpg
electron@diynoMacBook-Pro pytorch-openpose %

骨格は推定できなかった

この名画面の骨格推定は成功した

スクリーンショット 2021-07-19 21.27.48.png

Terminal
electron@diynoMacBook-Pro pytorch-openpose % ls images/moon_walker.jpg             
images/moon_walker.jpg
electron@diynoMacBook-Pro pytorch-openpose %

moon_walker.jpg

Terminal
electron@diynoMacBook-Pro pytorch-openpose % python3 test2.py ./images/moon_walker.jpg 
electron@diynoMacBook-Pro pytorch-openpose % ls result/pose_moon_walker.jpg 
result/pose_moon_walker.jpg
electron@diynoMacBook-Pro pytorch-openpose %

スクリーンショット 2021-07-19 21.32.21.png

0
1
0

Register as a new user and use Qiita more conveniently

  1. You get articles that match your needs
  2. You can efficiently read back useful information
  3. You can use dark theme
What you can do with signing up
Sign upLogin
0
1