2D Pose人体关键点检测(Python/Android /C++ Demo)

2022-05-24 04:14:43

2D Pose人体关键点检测(Python/Android /C++ Demo)

1.人体关键点数据集

（1）COCO数据集

（2）MPII数据集

（3）关键点示意图

2.人体关键点Demo（Python版本）

3.人体关键点Demo（C++版本）

4.人体关键点Demo（Android版本）

5.人体关键点检测效果

人体关键点检测（Human Keypoints Detection）又称为人体姿态估计，是计算机视觉中一个相对基础的任务，是人体动作识别、行为分析、人机交互等的前置任务。一般情况下可以将人体关键点检测细分为单人/多人关键点检测、2D/3D关键点检测，同时有算法在完成关键点检测之后还会进行关键点的跟踪，也被称为人体姿态跟踪。

本博客提供2D Pose的Python代码，以及C++版本的推理代码，还提供Android Demo APP，Android Demo 已经集成了轻量化版本的人体检测模型和人体姿态估计模型，在普通手机可实时检测。

Repo推荐：

个人repo(Python版): https://github.com/PanJinquan/Human-Keypoints-Detection
个人repo(C++版):https://github.com/PanJinquan/Human-Pose-Estimation-Lite-cpp
个人repo(Android版 支持CPU/GPU加速):https://github.com/PanJinquan/Human-Pose-Estimation-Lite-cpp/tree/master/data/app
《Simple Baselines for Human Pose Estimation and Tracking》https://github.com/Microsoft/human-pose-estimation.pytorch
《Deep High-Resolution Representation Learning for Human Pose Estimation》https://github.com/leoxiaobin/deep-high-resolution-net.pytorch

1.人体关键点数据集

（1）COCO数据集

下载地址：http://cocodataset.org/

COCO人体关键点标注，最多标注全身的17个关键点，平均一幅图像2个人，最多有13个人；

人体关键点标注，每个人体关键点个数的分布情况，其中11-15这个范围的人体是最多的，有接近70000人，6-10其次，超过40000人，后面依次为16-17,2-5,1.

（2）MPII数据集

下载地址：http://human-pose.mpi-inf.mpg.de/#download

人体关键点标注了全身16个关键点及其是否可见的信息，人数：train有28821，test有11701，有409种人类活动；使用mat的struct格式；行人框使用center和scale标注，人体尺度关于200像素高度。也就是除过了200

（3）关键点示意图

数据集	关键点示意图	关键点说明
COCO		# 图像左右翻转时，成对的关键点（训练时用于数据增强） flip_pairs=[[1, 2], [3, 4], [5, 6], [7, 8],[9, 10], [11, 12], [13, 14], [15, 16]] # 关键点连接线序号（用于绘制图像） skeleton =[[15, 13], [13, 11], [16, 14], [14, 12], [11, 12], [5, 11], [6, 12], [5, 6], [5, 7], [6, 8], [7, 9], [8, 10], [0, 1], [0, 2], [1, 3], [2, 4]] # 每个关键点序号对应人体关键点的意义 "keypoints": { 0: "nose", 1: "left_eye", 2: "right_eye", 3: "left_ear", 4: "right_ear", 5: "left_shoulder", 6: "right_shoulder", 7: "left_elbow", 8: "right_elbow", 9: "left_wrist", 10: "right_wrist", 11: "left_hip", 12: "right_hip", 13: "left_knee", 14: "right_knee", 15: "left_ankle", 16: "right_ankle" }
MPII		# 图像左右翻转时，成对的关键点（训练时用于数据增强） flip_pairs = [[0, 5], [1, 4], [2, 3], [10, 15], [11, 14], [12, 13]] # 关键点连接线序号（用于绘制图像） skeleton=[[0, 1], [1, 2], [3, 4], [4, 5], [2, 6], [6, 3], [12, 11], [7, 12], [11, 10], [13, 14], [14, 15], [8, 9], [8, 7], [6, 7], [7, 13]] # 每个关键点序号对应人体关键点的意义 "keypoints": {0: "r_ankle", 1: "r_knee", 2: "r_hip", 3: "l_hip", 4: "l_knee", 5: "l_ankle", 6: "pelvis", 7: "thorax", 8: "upper_neck", 9: "head_top", 10: " r_wrist", 11: "r_elbow", 12: "r_shoulder", 13: "l_shoulder", 14: "l_elbow", 15: "l_wrist" }
human3.6M
kinect		JointType_SpineBase = 0 # 脊柱底 JointType_SpineMid = 1 # 脊柱中间 JointType_Neck = 2 # 脖子 JointType_Head = 3 # 额头 JointType_ShoulderLeft = 4 JointType_ElbowLeft = 5 JointType_WristLeft = 6 JointType_HandLeft = 7 JointType_ShoulderRight = 8 JointType_ElbowRight = 9 JointType_WristRight = 10 JointType_HandRight = 11 JointType_HipLeft = 12 JointType_KneeLeft = 13 JointType_AnkleLeft = 14 JointType_FootLeft = 15 JointType_HipRight = 16 JointType_KneeRight = 17 JointType_AnkleRight = 18 JointType_FootRight = 19 JointType_SpineShoulder = 20 JointType_HandTipLeft = 21 JointType_ThumbLeft = 22 JointType_HandTipRight = 23 JointType_ThumbRight = 24 JointType_Count = 25