mediapipe教程1：python代码

一、背景

因为领导想用这个东西，因此开始研究，首先对于mediapipe的介绍不多说，百度一大堆，

主要是（1）跨平台，（2）对于移动设备友好，实时性足够

二、环境

因为对这东西不了解，这里先用python代码简单的看看效果，代码的参考网址：

（1）https://blog.csdn.net/javastart/article/details/116423090

（2）https://google.github.io/mediapipe/solutions/holistic

例子中只用到了cv2和mediapipe两个库，对于库安装下载不下来的可以考虑换源，参考：https://blog.csdn.net/afeiqiang/article/details/108257584

三、代码（没啥好说的，直接上代码，看代码即可）

（1）全身检测（图片）

import cv2
import mediapipe as mp

#mp.solutions.drawing_utils用于绘制
mp_drawing = mp.solutions.drawing_utils

#参数：1、颜色，2、线条粗细，3、点的半径
DrawingSpec_point = mp_drawing.DrawingSpec((0, 255, 0), 2, 2)
DrawingSpec_line = mp_drawing.DrawingSpec((0, 0, 255), 2, 2)

#mp.solutions.holistic是一个类别，是人的整体
mp_holistic = mp.solutions.holistic

#参数：1、是否检测静态图片，2、姿态模型的复杂度，3、结果看起来平滑（用于video有效），4、检测阈值，5、跟踪阈值
holistic = mp_holistic.Holistic(static_image_mode=True)

file = 'input.jpg'
image = cv2.imread(file)
image_hight, image_width, _ = image.shape
image1 = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

#处理RGB图像
results = holistic.process(image1)

#某一个点的坐标
if results.pose_landmarks:
  print(
      f'Nose coordinates: ('
      f'{results.pose_landmarks.landmark[mp_holistic.PoseLandmark.NOSE].x * image_width}, '
      f'{results.pose_landmarks.landmark[mp_holistic.PoseLandmark.NOSE].y * image_hight})'
  )

'''
mp_holistic.PoseLandmark类中共33个人体骨骼点
mp_holistic.HandLandmark类中共21个手部关键点
脸部有468个关键点
'''

#绘制
mp_drawing.draw_landmarks(
    image, results.face_landmarks, mp_holistic.FACE_CONNECTIONS, DrawingSpec_point, DrawingSpec_line)
mp_drawing.draw_landmarks(
    image, results.left_hand_landmarks, mp_holistic.HAND_CONNECTIONS, DrawingSpec_point, DrawingSpec_line)
mp_drawing.draw_landmarks(
    image, results.right_hand_landmarks, mp_holistic.HAND_CONNECTIONS, DrawingSpec_point, DrawingSpec_line)
mp_drawing.draw_landmarks(
    image, results.pose_landmarks, mp_holistic.POSE_CONNECTIONS, DrawingSpec_point, DrawingSpec_line)

#cv2.imshow('image',image)
#cv2.waitKey(0)
cv2.imwrite('image-holistic.jpg', image)
holistic.close()

（2）全身检测（视频）

import cv2
import mediapipe as mp

#mp.solutions.drawing_utils用于绘制
mp_drawing = mp.solutions.drawing_utils

#参数：1、颜色，2、线条粗细，3、点的半径
DrawingSpec_point = mp_drawing.DrawingSpec((0, 255, 0), 1 , 1)
DrawingSpec_line = mp_drawing.DrawingSpec((0, 0, 255), 1, 1)

#mp.solutions.holistic是一个类别，是人的整体
mp_holistic = mp.solutions.holistic

#参数：1、是否检测静态图片，2、姿态模型的复杂度，3、结果看起来平滑（用于video有效），4、检测阈值，5、跟踪阈值
holistic = mp_holistic.Holistic(min_detection_confidence=0.5, min_tracking_confidence=0.5)

cap = cv2.VideoCapture('input.mp4')
while cap.isOpened():
    success, image = cap.read()
    if not success:
        print("Ignoring empty camera frame.")
        continue
    image1 = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

    # 处理RGB图像
    results = holistic.process(image1)

    '''
    mp_holistic.PoseLandmark类中共33个人体骨骼点
    mp_holistic.HandLandmark类中共21个手部关键点
    脸部有468个关键点
    '''

    # 绘制
    mp_drawing.draw_landmarks(
        image, results.face_landmarks, mp_holistic.FACE_CONNECTIONS, DrawingSpec_point, DrawingSpec_line)
    mp_drawing.draw_landmarks(
        image, results.left_hand_landmarks, mp_holistic.HAND_CONNECTIONS, DrawingSpec_point, DrawingSpec_line)
    mp_drawing.draw_landmarks(
        image, results.right_hand_landmarks, mp_holistic.HAND_CONNECTIONS, DrawingSpec_point, DrawingSpec_line)
    mp_drawing.draw_landmarks(
        image, results.pose_landmarks, mp_holistic.POSE_CONNECTIONS, DrawingSpec_point, DrawingSpec_line)

    cv2.imshow('MediaPipe Holistic', image)
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

holistic.close()
cv2.destroyAllWindows()
cap.release()

（3）人体骨架检测（图片）

import cv2
import mediapipe as mp

#mp.solutions.drawing_utils用于绘制
mp_drawing = mp.solutions.drawing_utils

#参数：1、颜色，2、线条粗细，3、点的半径
DrawingSpec_point = mp_drawing.DrawingSpec((0, 255, 0), 2, 2)
DrawingSpec_line = mp_drawing.DrawingSpec((0, 0, 255), 2, 2)

#mp.solutions.pose，是人的骨架
mp_pose = mp.solutions.pose

#参数：1、是否检测静态图片，2、姿态模型的复杂度，3、结果看起来平滑（用于video有效），4、检测阈值，5、跟踪阈值
pose_mode = mp_pose.Pose(static_image_mode=True)

file = 'input.jpg'
image = cv2.imread(file)
image_hight, image_width, _ = image.shape
image1 = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

# 处理RGB图像
results = pose_mode.process(image1)

'''
mp_pose.PoseLandmark类中共33个人体骨骼点
'''
if results.pose_landmarks:
  print(
      f'Nose coordinates: ('
      f'{results.pose_landmarks.landmark[mp_pose.PoseLandmark.NOSE].x * image_width}, '
      f'{results.pose_landmarks.landmark[mp_pose.PoseLandmark.NOSE].y * image_hight})'
  )

#绘制
mp_drawing.draw_landmarks(
    image, results.pose_landmarks, mp_pose.POSE_CONNECTIONS, DrawingSpec_point, DrawingSpec_line)

#cv2.imshow('image',image)
#cv2.waitKey(0)
cv2.imwrite('image-pose.jpg', image)
pose_mode.close()

（4）人体骨架检测（视频）

import cv2
import mediapipe as mp

#mp.solutions.drawing_utils用于绘制
mp_drawing = mp.solutions.drawing_utils

#参数：1、颜色，2、线条粗细，3、点的半径
DrawingSpec_point = mp_drawing.DrawingSpec((0, 255, 0), 1 , 1)
DrawingSpec_line = mp_drawing.DrawingSpec((0, 0, 255), 1, 1)

#mp.solutions.pose，是人的骨架
mp_pose = mp.solutions.pose

#参数：1、是否检测静态图片，2、姿态模型的复杂度，3、结果看起来平滑（用于video有效），4、检测阈值，5、跟踪阈值
pose_mode = mp_pose.Pose(min_detection_confidence=0.5, min_tracking_confidence=0.5)

cap = cv2.VideoCapture('input.mp4')
while cap.isOpened():
    success, image = cap.read()
    if not success:
        print("Ignoring empty camera frame.")
        continue
    image1 = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

    # 处理RGB图像
    results = pose_mode.process(image1)

    '''
    mp_holistic.PoseLandmark类中共33个人体骨骼点
    '''

    # 绘制
    mp_drawing.draw_landmarks(
        image, results.pose_landmarks, mp_pose.POSE_CONNECTIONS, DrawingSpec_point, DrawingSpec_line)

    cv2.imshow('MediaPipe Pose', image)
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

pose_mode.close()
cv2.destroyAllWindows()
cap.release()

（5）手部关键点检测（图片）

import cv2
import mediapipe as mp

#mp.solutions.drawing_utils用于绘制
mp_drawing = mp.solutions.drawing_utils

#参数：1、颜色，2、线条粗细，3、点的半径
DrawingSpec_point = mp_drawing.DrawingSpec((0, 255, 0), 2, 2)
DrawingSpec_line = mp_drawing.DrawingSpec((0, 0, 255), 2, 2)

#mp.solutions.hands，是人的手
mp_hands = mp.solutions.hands

#参数：1、是否检测静态图片，2、手的数量，3、检测阈值，4、跟踪阈值
hands_mode = mp_hands.Hands(static_image_mode=True, max_num_hands=2)

file = 'input.jpg'
image = cv2.imread(file)
image_hight, image_width, _ = image.shape
image1 = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

#处理RGB图像
results = hands_mode.process(image1)

print('Handedness:', results.multi_handedness)
for hand_landmarks in results.multi_hand_landmarks:
    print('hand_landmarks:', hand_landmarks)
    print(
        f'Index finger tip coordinates: (',
        f'{hand_landmarks.landmark[mp_hands.HandLandmark.INDEX_FINGER_TIP].x * image_width}, '
        f'{hand_landmarks.landmark[mp_hands.HandLandmark.INDEX_FINGER_TIP].y * image_hight})'
    )
    mp_drawing.draw_landmarks(
        image, hand_landmarks, mp_hands.HAND_CONNECTIONS, DrawingSpec_point, DrawingSpec_line)

#cv2.imshow('image',image)
#cv2.waitKey(0)
cv2.imwrite('image-hands.jpg', image)
hands_mode.close()

（6）手部关键点检测（视频）

import cv2
import mediapipe as mp

#mp.solutions.drawing_utils用于绘制
mp_drawing = mp.solutions.drawing_utils

#参数：1、颜色，2、线条粗细，3、点的半径
DrawingSpec_point = mp_drawing.DrawingSpec((0, 255, 0), 1 , 1)
DrawingSpec_line = mp_drawing.DrawingSpec((0, 0, 255), 1, 1)

#mp.solutions.hands，是人的手
mp_hands = mp.solutions.hands

#参数：1、是否检测静态图片，2、手的数量，3、检测阈值，4、跟踪阈值
hands_mode = mp_hands.Hands(max_num_hands=2)

cap = cv2.VideoCapture('input.mp4')
while cap.isOpened():
    success, image = cap.read()
    if not success:
        print("Ignoring empty camera frame.")
        continue
    image1 = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

    # 处理RGB图像
    results = hands_mode.process(image1)

    # 绘制
    if results.multi_hand_landmarks:
        for hand_landmarks in results.multi_hand_landmarks:
            mp_drawing.draw_landmarks(
                image, hand_landmarks, mp_hands.HAND_CONNECTIONS, DrawingSpec_point, DrawingSpec_line)

    cv2.imshow('MediaPipe Hands', image)
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

hands_mode.close()
cv2.destroyAllWindows()
cap.release()