MediaPipe 是一款由 Google Research 开发并开源的多媒体机器学习模型应用框架。在谷歌，一系列重要产品，如、Google Lens、ARCore、Google Home 以及，都已深度整合了 MediaPipe。本文将介绍的为基于mediapipe的人体骨架提取方案。

1、mediapipe的安装

安装指令如下：

pip install mediapipe

官网地址：https://developers.google.cn/mediapipe
如果需要除了人体骨架提取以外的mediapipe的功能，可以参照官网内的demo进行编写。

2、demo编写

参照官网给的demo进行简要的更改，如下是对视频进行骨架提取，可根据需求更改为摄像头摄像或者照片。

import cv2
import time
import mediapipe as mp

mp_drawing = mp.solutions.drawing_utils
mp_drawing_styles = mp.solutions.drawing_styles
mp_holistic = mp.solutions.holistic

cap = cv2.VideoCapture('1_demo2.mp4')  # 替换为视频路径
fps_start_time = time.time()
fps = 0
with mp_holistic.Holistic(
        min_detection_confidence=0.5,
        min_tracking_confidence=0.5) as holistic:
    while cap.isOpened():
        success, image = cap.read()
        if not success:
            print("Video was ended.")
            break

        image.flags.writeable = False
        image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
        results = holistic.process(image)
        # 画图
        image.flags.writeable = True
        image = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)
        # 脸部骨架绘制
        mp_drawing.draw_landmarks(
            image,
            results.face_landmarks,
            mp_holistic.FACEMESH_CONTOURS,
            landmark_drawing_spec=None,
            connection_drawing_spec=mp_drawing_styles
                .get_default_face_mesh_contours_style())
        # 姿势绘制
        mp_drawing.draw_landmarks(
            image,
            results.pose_landmarks,
            mp_holistic.POSE_CONNECTIONS,
            landmark_drawing_spec=mp_drawing_styles
                .get_default_pose_landmarks_style())
        # 左右手绘制
        mp_drawing.draw_landmarks(image, results.left_hand_landmarks, mp_holistic.HAND_CONNECTIONS)
        mp_drawing.draw_landmarks(image, results.right_hand_landmarks, mp_holistic.HAND_CONNECTIONS)
        cv2.imshow('MediaPipe Holistic', cv2.flip(image, 1))
        fps_end_time = time.time()
        time_diff = fps_end_time - fps_start_time
        if time_diff >= 1:
            fps = int(1 / time_diff)
            fps_start_time = time.time()
        cv2.putText(image, f"FPS: {fps}", (10, 30), cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 255, 0), 2)
        if cv2.waitKey(5) & 0xFF == 27:
            break
# cap.release()
cv2.destroyAllWindows()

运行结果如下：