OpenPose人体姿态识别

OpenPose人体姿态识别项目是美国卡耐基梅隆大学（CMU）基于卷积神经网络和监督学习并以caffe为框架开发的开源库。可以实现人体动作、面部表情、手指运动等姿态估计。适用于单人和多人，具有极好的鲁棒性。是世界上首个基于深度学习的实时多人二维姿态估计应用，基于它的实例如雨后春笋般涌现。人体姿态估计技术在体育健身、动作采集、3D试衣、舆情监测等领域具有广阔的应用前景，人们更加熟悉的应用就是抖音尬舞机。

此案例使用的是PyTorch版本

下载代码和模型

!wget https://obs-aigallery-zc.obs.cn-north-4.myhuaweicloud.com/algorithm/openpose.zip
!unzip openpose.zip

检测人体图像

!pip install ffmpeg-python
!pip install imageio
import sys
sys.path.insert(0, './openpose/src/')
import cv2
import matplotlib.pyplot as plt
import copy
import numpy as np

import model
import util
from body import Body
from hand import Hand


from glob import glob
import os
import argparse
import json

# video file processing setup
# from: https://stackoverflow.com/a/61927951

import subprocess
from pathlib import Path
from typing import NamedTuple


body_estimation = Body('./openpose/models/body_pose_model.pth')
hand_estimation = Hand('./openpose/models/hand_pose_model.pth')

test_image = './openpose/images/demo.jpg'
oriImg = cv2.imread(test_image)  # B,G,R order
candidate, subset = body_estimation(oriImg)
canvas = copy.deepcopy(oriImg)
canvas = util.draw_bodypose(canvas, candidate, subset)
# detect hand
hands_list = util.handDetect(candidate, subset, oriImg)

all_hand_peaks = []
for x, y, w, is_left in hands_list:
    # cv2.rectangle(canvas, (x, y), (x+w, y+w), (0, 255, 0), 2, lineType=cv2.LINE_AA)
    # cv2.putText(canvas, 'left' if is_left else 'right', (x, y), cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 0, 255), 2)

    # if is_left:
        # plt.imshow(oriImg[y:y+w, x:x+w, :][:, :, [2, 1, 0]])
        # plt.show()
    peaks = hand_estimation(oriImg[y:y+w, x:x+w, :])
    peaks[:, 0] = np.where(peaks[:, 0]==0, peaks[:, 0], peaks[:, 0]+x)
    peaks[:, 1] = np.where(peaks[:, 1]==0, peaks[:, 1], peaks[:, 1]+y)
    # else:
    #     peaks = hand_estimation(cv2.flip(oriImg[y:y+w, x:x+w, :], 1))
    #     peaks[:, 0] = np.where(peaks[:, 0]==0, peaks[:, 0], w-peaks[:, 0]-1+x)
    #     peaks[:, 1] = np.where(peaks[:, 1]==0, peaks[:, 1], peaks[:, 1]+y)
    #     print(peaks)
    all_hand_peaks.append(peaks)

canvas = util.draw_handpose(canvas, all_hand_peaks)

plt.imshow(canvas[:, :, [2, 1, 0]])
plt.axis('off')
plt.show()

检测人体视频

输入视频为’./openpose/images/play.mp4’，可替换成需要检测的

由于视频太长只处理了其中的10帧

输出gif图，路径output_file会打印

import imageio
class FFProbeResult(NamedTuple):
    return_code: int
    json: str
    error: str


def ffprobe(file_path) -> FFProbeResult:
    command_array = ["ffprobe",
                     "-v", "quiet",
                     "-print_format", "json",
                     "-show_format",
                     "-show_streams",
                     file_path]
    result = subprocess.run(command_array, stdout=subprocess.PIPE, stderr=subprocess.PIPE, universal_newlines=True)
    return FFProbeResult(return_code=result.returncode,
                         json=result.stdout,
                         error=result.stderr)


def process_frame(frame, body=True, hands=True):
    canvas = copy.deepcopy(frame)
    if body:
        candidate, subset = body_estimation(frame)
        canvas = util.draw_bodypose(canvas, candidate, subset)
    if hands:
        hands_list = util.handDetect(candidate, subset, frame)
        all_hand_peaks = []
        for x, y, w, is_left in hands_list:
            peaks = hand_estimation(frame[y:y+w, x:x+w, :])
            peaks[:, 0] = np.where(peaks[:, 0]==0, peaks[:, 0], peaks[:, 0]+x)
            peaks[:, 1] = np.where(peaks[:, 1]==0, peaks[:, 1], peaks[:, 1]+y)
            all_hand_peaks.append(peaks)
        canvas = util.draw_handpose(canvas, all_hand_peaks)
    return canvas

# writing video with ffmpeg because cv2 writer failed
# https://stackoverflow.com/questions/61036822/opencv-videowriter-produces-cant-find-starting-number-error
import ffmpeg

# open specified video
video_file = './openpose/images/play.mp4'
cap = cv2.VideoCapture(video_file)

# get video file info
ffprobe_result = ffprobe(video_file)
info = json.loads(ffprobe_result.json)
videoinfo = [i for i in info["streams"] if i["codec_type"] == "video"][0]
input_fps = videoinfo["avg_frame_rate"]
# input_fps = float(input_fps[0])/float(input_fps[1])
input_pix_fmt = videoinfo["pix_fmt"]
input_vcodec = videoinfo["codec_name"]

# define a writer object to write to a movidified file
output_file = ".".join(video_file.split(".")[:-1])+".processed.gif" 
print(output_file)

buff=[]
count = 0
while(cap.isOpened()):
    ret, frame = cap.read()
    if frame is None or count>10:  # 修改这里可调节视频处理帧数
        break

    posed_frame = process_frame(frame, body=True,
                                       hands=True)

    buff.append(posed_frame)
    count += 1

gif=imageio.mimsave(output_file,buff,'GIF')
cap.release()
cv2.destroyAllWindows()

./openpose/images/play.processed.gif