Mediapipe 仅显示车身地标

Question

我已经在 Windows 11 上使用 Python (3.7.0) 安装了 Mediapipe (0.9.0.1)。我已经能够成功让 Mediapipe 生成地标（用于面部和身体）；用于图像、视频和网络摄像头流。

我现在想让 Mediapipe仅绘制身体特定的标志（即排除面部标志）。

我知道我可以使用 OpenCV （或 Czone）来实现这个目标，但是，我希望使用 Mediapipe （即使用draw_landmarksMediaPipe 库中的函数）来实现我的目标。

我正在尝试的具体代码（但有错误）如下：

#Initialize a list to store the detected landmarks.
    landmarks = []

    # Iterate over the Mediapipe detected landmarks.
    for landmark in results.pose_landmarks.landmark:
        
        # Append the Mediapipe landmark into the list.
        landmarks.append((int(landmark.x * width), int(landmark.y * height),
                                (landmark.z * width)))
    
    #create index list for specific landmarks 
    body_landmark_indices = [11,12,13,14,15,16,23,24,25,26,27,28,29,30,31,32]

    landmark_list_body = []

    #Create a list which only has the required landmarks
    for index in body_landmark_indices:
        landmark_list_body.append(landmarks[index - 1])

    mp_drawing.draw_landmarks(
                        image=output_image,
                        landmark_list=landmark_list_body.pose_landmarks,
                        connections=mp_pose.POSE_CONNECTIONS,
                        landmark_drawing_spec=landmark_drawing_spec,
                        connection_drawing_spec=connection_drawing_spec)`

Executing the above I get the error `'list' object has no attribute 'pose_landmarks'

我已替换landmark_list=landmark_list_body.pose_landmarks,为landmark_list=landmark_list_body但有错误。

我现在已经非常分层并且没有想法了。世界上有没有披风的英雄吗？

谢谢。

Answer 1

您可以尝试以下方法：

import cv2
import mediapipe as mp
import numpy as np
from mediapipe.python.solutions.pose import PoseLandmark
from mediapipe.python.solutions.drawing_utils import DrawingSpec
mp_drawing = mp.solutions.drawing_utils
mp_drawing_styles = mp.solutions.drawing_styles
mp_pose = mp.solutions.pose


custom_style = mp_drawing_styles.get_default_pose_landmarks_style()
custom_connections = list(mp_pose.POSE_CONNECTIONS)

# list of landmarks to exclude from the drawing
excluded_landmarks = [
    PoseLandmark.LEFT_EYE, 
    PoseLandmark.RIGHT_EYE, 
    PoseLandmark.LEFT_EYE_INNER, 
    PoseLandmark.RIGHT_EYE_INNER, 
    PoseLandmark.LEFT_EAR,
    PoseLandmark.RIGHT_EAR,
    PoseLandmark.LEFT_EYE_OUTER,
    PoseLandmark.RIGHT_EYE_OUTER,
    PoseLandmark.NOSE,
    PoseLandmark.MOUTH_LEFT,
    PoseLandmark.MOUTH_RIGHT ]

for landmark in excluded_landmarks:
    # we change the way the excluded landmarks are drawn
    custom_style[landmark] = DrawingSpec(color=(255,255,0), thickness=None) 
    # we remove all connections which contain these landmarks
    custom_connections = [connection_tuple for connection_tuple in custom_connections 
                            if landmark.value not in connection_tuple]


IMAGE_FILES = ["test.jpg"]
BG_COLOR = (192, 192, 192) 
with mp_pose.Pose(
    static_image_mode=True,
    model_complexity=2,
    enable_segmentation=True,
    min_detection_confidence=0.5) as pose:
    for idx, file in enumerate(IMAGE_FILES):
        image = cv2.imread(file)
        image_height, image_width, _ = image.shape
        results = pose.process(cv2.cvtColor(image, cv2.COLOR_BGR2RGB))
        annotated_image = image.copy()

        mp_drawing.draw_landmarks(
            annotated_image,
            results.pose_landmarks,
            connections = custom_connections, #  passing the modified connections list
            landmark_drawing_spec=custom_style) # and drawing style 

        cv2.imshow('landmarks', annotated_image)
        cv2.waitKey(0)

它修改DrawingSpec和POSE_CONNECTIONS以“隐藏”地标的子集。但是，由于该draw_landmarks()功能在 Mediapipe 中的实现方式，还需要在drawing_utils.py（位于site-packages/mediapipe/python/solutions）中添加一个条件：

if drawing_spec.thickness == None: continue

将其添加到第 188 行之前（请参阅MediaPipe GitHub 存储库, # White circle border）。结果应该是这样的：

...
      drawing_spec = landmark_drawing_spec[idx] if isinstance(
          landmark_drawing_spec, Mapping) else landmark_drawing_spec
      if drawing_spec.thickness == None: continue
      # White circle border
      circle_border_radius = max(drawing_spec.circle_radius + 1,
                                 int(drawing_spec.circle_radius * 1.2))
...

需要进行此更改，以便完全消除地标周围绘制的白色边框，无论其绘图规范如何。

希望能帮助到你。