MediaPipe人体姿态、手指关键点检测

一缕清波

MediaPipe人体姿态、手指关键点检测文章目录MediaPipe人体姿态、手指关键点检测前言一、手指关键点检测二、姿态检测三、3D物体案例检测案例前言  Mediapipe是google的一个开源项目，用于构建机器学习管道。  提供了16个预训练模型的案例：人脸检测、FaceMesh、虹膜、手、姿态、人体、人物分割、头发分割、目标检测、BoxTracking、InstantMotionTracking、3D目标检测、特征匹配、AutoFlip、MediaSequence、YouTube-8M。  肢体识别本质上还是分类任务，该技术有很多应用场景，比如手势识别控制类应用、动作检测类应用、动作评测类应用、以及一些移动设备AR视频合成类应用。一、手指关键点检测检测一只手中21个关键点坐标，每个点对应一个名称importcv2importmediapipeasmp#用来在图片中画出关键点mp_drawing=mp.solutions.drawing_utils#关键点图样式mp_drawing_styles=mp.solutions.drawing_stylesmp_hands=mp.solutions.hands#MAX_NUM_HANDS要检测的最大手数默认为2hands=mp_hands.Hands(max_num_hands=1)if__name__=='__main__':file='1.png'#图片翻转image=cv2.flip(cv2.imread(file),1)results=hands.process(cv2.cvtColor(image,cv2.COLOR_BGR2RGB))image_height,image_width,_=image.shapeannotated_image=image.copy()#遍历检测出来的手掌forhand_landmarksinresults.multi_hand_landmarks:forlminhand_landmarks.landmark:#获取每个点的坐标x=lm.x*image_widthy=lm.y*image_height#画关键点mp_drawing.draw_landmarks(annotated_image,hand_landmarks,mp_hands.HAND_CONNECTIONS,mp_drawing_styles.get_default_hand_landmarks_style(),mp_drawing_styles.get_default_hand_connections_style())cv2.imwrite('0.png',cv2.flip(annotated_image,1))12345678910111213141516171819202122232425262728293031323334二、姿态检测importcv2importmediapipeasmp#用来在图片中画出关键点mp_drawing=mp.solutions.drawing_utils#关键点图样式mp_drawing_styles=mp.solutions.drawing_stylesmpPose=mp.solutions.posepose=mpPose.Pose()if__name__=='__main__':file='1.png'image=cv2.imread(file)image_height,image_width,_=image.shapeimgRGB=cv2.cvtColor(image,cv2.COLOR_BGR2RGB)results=pose.process(imgRGB)#mediapipe姿态检测只能检测一个人ifresults.pose_landmarks:forlminresults.pose_landmarks.landmark:h,w,c=image.shapecx,cy=int(lm.x*w),int(lm.y*h)#画关键点annotated_image=image.copy()mp_drawing.draw_landmarks(annotated_image,results.pose_landmarks,mpPose.POSE_CONNECTIONS,landmark_drawing_spec=mp_drawing_styles.get_default_pose_landmarks_style())cv2.imwrite('0.png',annotated_image)12345678910111213141516171819202122232425262728293031三、3D物体案例检测案例其他案例具体可以查看官网详细介绍：https://google.github.io/mediapipe/solutions/pose.html