mediapipe像素坐标转换

Mediapipe中的像素坐标转换可以使用以下代码实现：

import mediapipe as mp

mp_drawing = mp.solutions.drawing_utils

# 定义一个坐标转换函数
def pixel2normalized(pixel_coordinates, image_width, image_height):
    """将像素坐标转化为归一化的坐标"""
    x, y = pixel_coordinates
    return (x / image_width, y / image_height)

# 使用MediaPipe获取手部关键点坐标
with mp.solutions.hands.Hands(static_image_mode=True, max_num_hands=1, min_detection_confidence=0.5) as hands:
    image = cv2.imread('image.jpg')
    results = hands.process(cv2.cvtColor(image, cv2.COLOR_BGR2RGB))
    if results.multi_hand_landmarks:
        for hand_landmarks in results.multi_hand_landmarks:
            # 将关键点坐标从像素坐标系转换为归一化坐标系
            normalized_landmarks = [pixel2normalized((lm.x, lm.y), image.shape[1], image.shape[0]) for lm in hand_landmarks.landmark]
            # 将归一化坐标系转换为像素坐标系
            pixel_landmarks = [mp_drawing._normalized_to_pixel_coordinates(lm.x, lm.y, image.shape[1], image.shape[0]) for lm in normalized_landmarks]
            # 在图像中绘制转换后的关键点
            mp_drawing.draw_landmarks(image, hand_landmarks, mp_hands.HAND_CONNECTIONS)

在上述代码中，`pixel2normalized()`函数将像素坐标系下的坐标转换为归一化坐标系下的坐标，`mp_drawing._normalized_to_pixel_coordinates()`函数则将归一化坐标系下的坐标转换为像素坐标系下的坐标。