目录

写在前面：

1、准备工作：

环境准备：

源码下载：

2、编译

3、测试：

测试视频：

测试图片：

 

写在前面：

OpenPose人体姿态识别项目是美国卡耐基梅隆大学（CMU）开发的开源库。代码开源已经有相当一段时间了，各位博主博客的经验也很丰富，但由于版本关系，新版本的部署方式和之前有些不同。

本文仅记录win7 下编译配置过程。

参考：【OpenPose-Windows】OpenPose+VS2015+Windows+CUDA8+cuDNN5.1 官方配置教程

 

1、准备工作：

环境准备：

VS2015 with update3
	
cuda10.0， 下载地址
	
cudnn7。下载地址  注意下载对应cuda的版本，下载后解压，把解压后的三个文件夹直接拷贝<cuda安装目录>Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\里，覆盖即可。
	
cmake 下载地址
源码下载：

openpose项目github主页： https://github.com/CMU-Perceptual-Computing-Lab/openpose

各发布版本：https://github.com/CMU-Perceptual-Computing-Lab/openpose/releases

我们从发布版本中下载，当前最新为V1.5.1,



目录里还有openpose-1.5.1-binaries-win64... zip等，是已经编译好的，如果不想自己编译，（边缘还是比较简单的）， 可以直接下载使用。

本文下载源码ZIP压缩包，下载后解压，解压目录 “E:\code2019\new_alg\OpenPose\”。

 

2、编译

打开cmake-gui，设置

源路径：“E:/code2019/new_alg/OpenPose/openpose-1.5.1”

生成路径：“E:/code2019/new_alg/OpenPose/openpose-1.5.1/windows”

点击config，编译器本文选了了" visual Studio 14 2015 win64"



这个过程时间有点长，会下载依赖的文件和检测需要的模型，之前的版本是手动使用脚本下载。

config完成后，跳出一个喜庆的大红的界面，重点看一下里面CUDA_开头的项。



我电脑上装了多个cuda版本，所以CUDA_ARCH修改为了手动，CUDA_TOOLKIT_ROOT_DIR指定路径为“C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0”。

再次点击config，完成后点击 Generate。

生成完毕后，打开生成的文件夹，里面bin目录里有运行环境所依赖的动态库，包括boost,caffe,cuda,cudnn等，通过文件名可以看出，默认cuda版本为10.0，cudnn版本为7。



用vs2015打开工程，生成ALL_BUILD。



编译完成后，先将<工程目录>\x64\Release文件夹内的所有文件拷贝到<工程目录>bin文件夹内。

然后将<工程目录>bin目录整个拷贝到外层文件夹，本文为“E:\code2019\new_alg\OpenPose\openpose-1.5.1”目录。测试用到了这个目录的“models”和“examples”文件夹内容。

注意一共拷贝了两次。

 

3、测试：

打开cmd.exe命令行，目录切换至“E:\code2019\new_alg\OpenPose\openpose-1.5.1”

测试视频：

bin\OpenPoseDemo.exe --video examples\media\video.avi

bin\OpenPoseDemo.exe --video examples\media\video.avi --face --hand

 



我的显卡是:Geforce gtx 745, 只检测关键点，帧率1.4

 

换个比较好GPU的服务器运行,速度提升还是很明显的，帧率达到了12帧



 

测试图片：

bin\OpenPoseDemo.exe --image_dir examples\media\

bin\OpenPoseDemo.exe --image_dir examples\media\ --face --hand

 



 

转载于：

PS：1、请先按照步骤安装好依赖库，否则后续报错会很折磨。

2、最后openpose根目录下新建的build中make编译成功后，请在根目录下运行测试指令，不然会报错。

压缩包 : a3e7eacaf0843d216b9d53a2f46b121.zip 列表
tf-pose-estimation-master/
tf-pose-estimation-master/.dockerignore
tf-pose-estimation-master/.gitattributes
tf-pose-estimation-master/.gitignore
tf-pose-estimation-master/CMakeLists.txt
tf-pose-estimation-master/Dockerfile
tf-pose-estimation-master/LICENSE
tf-pose-estimation-master/README.md
tf-pose-estimation-master/__init__.py
tf-pose-estimation-master/docker/
tf-pose-estimation-master/docker/Dockerfile
tf-pose-estimation-master/docker/README.md
tf-pose-estimation-master/etcs/
tf-pose-estimation-master/etcs/dance.mp4
tf-pose-estimation-master/etcs/experiments.md
tf-pose-estimation-master/etcs/feature.md
tf-pose-estimation-master/etcs/imgcat0.gif
tf-pose-estimation-master/etcs/inference_result2.png
tf-pose-estimation-master/etcs/loss_ll_heat.png
tf-pose-estimation-master/etcs/loss_ll_paf.png
tf-pose-estimation-master/etcs/openpose_macbook13_mobilenet2.gif
tf-pose-estimation-master/etcs/openpose_macbook_cmu.gif
tf-pose-estimation-master/etcs/openpose_macbook_mobilenet3.gif
tf-pose-estimation-master/etcs/openpose_p40_cmu.gif
tf-pose-estimation-master/etcs/openpose_p40_mobilenet.gif
tf-pose-estimation-master/etcs/openpose_tx2_mobilenet3.gif
tf-pose-estimation-master/etcs/ros.md
tf-pose-estimation-master/etcs/training.md
tf-pose-estimation-master/images/
tf-pose-estimation-master/images/COCO_val2014_000000000357.jpg
tf-pose-estimation-master/images/apink1.jpg
tf-pose-estimation-master/images/apink1_crop.jpg
tf-pose-estimation-master/images/apink1_crop_s1.jpg
tf-pose-estimation-master/images/apink2.jpg
tf-pose-estimation-master/images/apink3.jpg
tf-pose-estimation-master/images/cat.jpg
tf-pose-estimation-master/images/golf.jpg
tf-pose-estimation-master/images/hand1.jpg
tf-pose-estimation-master/images/hand1_small.jpg
tf-pose-estimation-master/images/hand2.jpg
tf-pose-estimation-master/images/handsup1.jpg
tf-pose-estimation-master/images/p1.jpg
tf-pose-estimation-master/images/p2.jpg
tf-pose-estimation-master/images/p3.jpg
tf-pose-estimation-master/images/p3_dance.png
tf-pose-estimation-master/images/ski.jpg
tf-pose-estimation-master/images/valid_person1.jpg
tf-pose-estimation-master/launch/
tf-pose-estimation-master/launch/demo_video.launch
tf-pose-estimation-master/models/
tf-pose-estimation-master/models/__init__.py
tf-pose-estimation-master/models/graph/
tf-pose-estimation-master/models/graph/cmu/
tf-pose-estimation-master/models/graph/cmu/__init__.py
tf-pose-estimation-master/models/graph/cmu/download.sh
tf-pose-estimation-master/models/graph/mobilenet_thin/
tf-pose-estimation-master/models/graph/mobilenet_thin/__init__.py
tf-pose-estimation-master/models/graph/mobilenet_thin/graph.pb
tf-pose-estimation-master/models/graph/mobilenet_thin/graph_freeze.pb
tf-pose-estimation-master/models/graph/mobilenet_thin/graph_opt.pb
tf-pose-estimation-master/models/numpy/
tf-pose-estimation-master/models/numpy/download.sh
tf-pose-estimation-master/models/pretrained/
tf-pose-estimation-master/models/pretrained/mobilenet_v1_0.50_224_2017_06_14/
tf-pose-estimation-master/models/pretrained/mobilenet_v1_0.50_224_2017_06_14/download.sh
tf-pose-estimation-master/models/pretrained/mobilenet_v1_0.75_224_2017_06_14/
tf-pose-estimation-master/models/pretrained/mobilenet_v1_0.75_224_2017_06_14/download.sh
tf-pose-estimation-master/models/pretrained/mobilenet_v1_1.0_224_2017_06_14/
tf-pose-estimation-master/models/pretrained/mobilenet_v1_1.0_224_2017_06_14/download.sh
tf-pose-estimation-master/models/pretrained/resnet_v2_101/
tf-pose-estimation-master/models/pretrained/resnet_v2_101/download.sh
tf-pose-estimation-master/msg/
tf-pose-estimation-master/msg/BodyPartElm.msg
tf-pose-estimation-master/msg/Person.msg
tf-pose-estimation-master/msg/Persons.msg
tf-pose-estimation-master/package.xml
tf-pose-estimation-master/requirements.txt
tf-pose-estimation-master/run.py
tf-pose-estimation-master/run_checkpoint.py
tf-pose-estimation-master/run_directory.py
tf-pose-estimation-master/run_video.py
tf-pose-estimation-master/run_webcam.py
tf-pose-estimation-master/scripts/
tf-pose-estimation-master/scripts/broadcaster_ros.py
tf-pose-estimation-master/scripts/visualization.py
tf-pose-estimation-master/setup.py
tf-pose-estimation-master/tf_pose/
tf-pose-estimation-master/tf_pose/__init__.py
tf-pose-estimation-master/tf_pose/common.py
tf-pose-estimation-master/tf_pose/datum_pb2.py
tf-pose-estimation-master/tf_pose/estimator.py
tf-pose-estimation-master/tf_pose/eval.py
tf-pose-estimation-master/tf_pose/network_base.py
tf-pose-estimation-master/tf_pose/network_cmu.py
tf-pose-estimation-master/tf_pose/network_dsconv.py
tf-pose-estimation-master/tf_pose/network_mobilenet.py
tf-pose-estimation-master/tf_pose/network_mobilenet_thin.py
tf-pose-estimation-master/tf_pose/network_personlab.py
tf-pose-estimation-master/tf_pose/networks.py
tf-pose-estimation-master/tf_pose/pafprocess/
tf-pose-estimation-master/tf_pose/pafprocess/README.md
tf-pose-estimation-master/tf_pose/pafprocess/__init__.py
tf-pose-estimation-master/tf_pose/pafprocess/numpy.i
tf-pose-estimation-master/tf_pose/pafprocess/pafprocess.cpp
tf-pose-estimation-master/tf_pose/pafprocess/pafprocess.h
tf-pose-estimati

参考博客：Openpose驾驶员危险驾驶检测（抽烟打电话）
人体姿态识别模型—openpose
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。

参考：Github开源人体姿态识别项目OpenPose中文文档

OpenPose项目Github链接：https://github.com/CMU-Perceptual-Computing-Lab/openpose
openpose模型有：手，脸，人体姿态。如图： 
CoCo人体18点检测-图片
import cv2
import matplotlib.pyplot as plt
#配置文件
protoFile = './models/pose/coco/pose_deploy_linevec.prototxt'  
weightsfile = './models/pose/coco/pose_iter_440000.caffemodel' 
npoints = 18
POSE_PAIRS = [[1,0],[1,2],[1,5],[2,3],[3,4],[5,6],[6,7],[1,8],[8,9],[9,10],[1,11],[11,12],[12,13],[0,14],[0,15],[14,16],[15,17]]
#加载网络
net = cv2.dnn.readNetFromCaffe(protoFile,weightsfile)
#读取图片
im = cv2.imread('kai.jpg')
#（1440*1080）
im = cv2.cvtColor(im,cv2.COLOR_BGR2RGB)
inHeight = im.shape[0]  #1440
inWidth = im.shape[1]   #1080
netInputsize = (368,368)
#转为inpBlob格式
inpBlob = cv2.dnn.blobFromImage(im,1.0/255,netInputsize,(0,0,0),swapRB=True,crop=False)
#归一化后的图片最为输入传入net网络,然后输出
net.setInput(inpBlob)
output = net.forward()  #1*57*46*46


# plt.figure(figsize=(20,10))
# for i in range(npoints):
#     probMap = output[0,i,:,:]  #把18个output[2][3]输出出来  probMap.shape(46*46)
#     displayMap = cv2.resize(probMap,(inWidth,inHeight),cv2.INTER_LINEAR) #1440*1080
#     plt.subplot(3,6,i+1);plt.axis('on');
#     plt.imshow(displayMap,cmap='jet')
# plt.show()


scaleX = float(inWidth) / output.shape[3]    #float(1080)/64 = 23.47826086956522
scaleY = float(inHeight)/ output.shape[2]    #float(1440)/64 = 31.304347826086957
points = []
threshold = 0.1
for i in range(npoints):
    probMap = output[0,i,:,:]  #shape(46*46)
    minVal,prob,minLoc,point =cv2.minMaxLoc(probMap)
    x = scaleX * point[0]
    y = scaleY * point[1]
    if prob > threshold:
        points.append((int(x),int(y)))
    else:
        points.append(None)
    #points[]最后为18个关键点的坐标
    #[(516, 313), (516, 438), (399, 438), (375, 626), (352, 751), (610, 438), (633, 594), (657, 751), (446, 751),
    # (446, 970), (446, 1158), (563, 782), (540, 1001), (540, 1064), (493, 281), (540, 281), (446, 313), (563, 313)]


imPoints = im.copy()
imSkeleton = im.copy()
for i,p in enumerate(points):
    #enumerate把points的值前面带上索引i
    cv2.circle(imPoints,p,8,(255,255,0),thickness=1,lineType=cv2.FILLED)
    cv2.putText(imPoints,'{}'.format(i),p,cv2.FONT_HERSHEY_SIMPLEX,1,(255,0,0),2,lineType=cv2.LINE_AA)
for pair in POSE_PAIRS:
    partA = pair[0]
    partB = pair[1]
    #if points[partA] and points[partB]:
    cv2.line(imSkeleton,points[partA],points[partB],(255, 255,0),2)
    cv2.circle(imSkeleton, points[partA],8,(255,0,0),thickness=-1, lineType=cv2.FILLED)

plt.figure(figsize=(20,10))
plt.subplot(121)
plt.axis('off');plt.imshow(imPoints)
plt.subplot(122)
plt.axis('off');plt.imshow(imSkeleton)
plt.show()
检测结果：


CoCo人体18点检测-视频

把上述代码改为由opencv捕获摄像头，实时进行人体姿态检测

import cv2
def detect(img):
    protoFile = './models/pose/coco/pose_deploy_linevec.prototxt'
    weightsfile = './models/pose/coco/pose_iter_440000.caffemodel'
    npoints = 18
    POSE_PAIRS = [[1, 0], [1, 2], [1, 5], [2, 3], [3, 4], [5, 6], [6, 7], [1, 8], [8, 9], [9, 10], [1, 11], [11, 12],
                  [12, 13], [0, 14], [0, 15], [14, 16], [15, 17]]
    net = cv2.dnn.readNetFromCaffe(protoFile, weightsfile)
    inHeight = img.shape[0]  # 1440
    inWidth = img.shape[1]  # 1080
    netInputsize = (368, 368)
    inpBlob = cv2.dnn.blobFromImage(img, 1.0 / 255, netInputsize, (0, 0, 0), swapRB=True, crop=False)
    net.setInput(inpBlob)
    output = net.forward()

    scaleX = float(inWidth) / output.shape[3]
    scaleY = float(inHeight) / output.shape[2]
    points = []
    threshold = 0.1
    for i in range(npoints):
        probMap = output[0, i, :, :]  # shape(46*46)
        minVal, prob, minLoc, point = cv2.minMaxLoc(probMap)
        x = scaleX * point[0]
        y = scaleY * point[1]
        if prob > threshold:
            points.append((int(x), int(y)))
        else:
            points.append(None)

    for i, p in enumerate(points):
        # enumerate把points的值前面带上索引i
        cv2.circle(img, p, 8, (255, 255, 0), thickness=1, lineType=cv2.FILLED)
        cv2.putText(img, '{}'.format(i), p, cv2.FONT_HERSHEY_SIMPLEX, 1, (255, 0, 0), 2, lineType=cv2.LINE_AA)


cap = cv2.VideoCapture(0)
while True:
    ret, frame = cap.read()
    detect(frame)

    cv2.imshow('frame',frame)
    if cv2.waitKey(1) & 0xFF == ord('q'):  # 按‘q’退出
        break
检测结果：


距离计算
为了检测驾驶员是否存在吸烟和打电话等危险行为，可以计算人体特定关键点之间的距离。

吸烟：右手到鼻子的距离或者左手到鼻子的距离。

打电话：右手到右耳的距离或者左手到左耳的距离。
import math
def __distance(A,B):
    if A is None or B is None:
        return 0
    else:
        return math.sqrt((A[0]-B[0])**2 + (A[1]-B[1])**2)

distance0 =  __distance(points[0],points[4]) #右手到鼻子距离
distance1 =  __distance(points[0],points[7]) #左手到鼻子距离
distance2 =  __distance(points[4],points[16]) #右手到右耳距离
distance3 =  __distance(points[7],points[17]) #左手到左耳距离
print('右手到鼻子距离：',distance0,'左手到鼻子距离：',distance1)
print('右手到右耳距离：',distance2,'左手到左耳距离:',distance3
计算结果
右手到鼻子距离： 467.6964827748868 左手到鼻子距离： 460.135849505339
右手到右耳距离： 447.97321348491363 左手到左耳距离: 447.97321348491363