[MediaPipe] Pose Landmark Detection 자세 특징 감지

MediaPipe를 사용해 자세 특징을 감지해 보자

 

import numpy as np
import cv2
from mediapipe import solutions
from mediapipe.framework.formats import landmark_pb2
import mediapipe as mp
from mediapipe.tasks import python
from mediapipe.tasks.python import vision
 
def draw_landmarks_on_image(rgb_image, detection_result):
  pose_landmarks_list = detection_result.pose_landmarks
  annotated_image = np.copy(rgb_image)
 
  # Loop through the detected poses to visualize.
  for idx in range(len(pose_landmarks_list)):
    pose_landmarks = pose_landmarks_list[idx]
 
    # Draw the pose landmarks.
    pose_landmarks_proto = landmark_pb2.NormalizedLandmarkList()
    pose_landmarks_proto.landmark.extend([
      landmark_pb2.NormalizedLandmark(x=landmark.x, y=landmark.y, z=landmark.z) for landmark in pose_landmarks
    ])
    solutions.drawing_utils.draw_landmarks(
      annotated_image,
      pose_landmarks_proto,
      solutions.pose.POSE_CONNECTIONS,
      solutions.drawing_styles.get_default_pose_landmarks_style())
  return annotated_image
 
# Create an PoseLandmarker object.
base_options = python.BaseOptions(model_asset_path='pose_landmarker_full.task')
# https://ai.google.dev/edge/mediapipe/solutions/vision/pose_landmarker
options = vision.PoseLandmarkerOptions(base_options=base_options, output_segmentation_masks=True)
detector = vision.PoseLandmarker.create_from_options(options)
 
# Load the input image.
image = mp.Image.create_from_file("pose.jpg")
 
# Detect pose landmarks from the input image.
detection_result = detector.detect(image)
 
# Process the detection result. In this case, visualize it.
annotated_image = draw_landmarks_on_image(image.numpy_view(), detection_result)
cv2.imshow('sean', cv2.cvtColor(annotated_image, cv2.COLOR_RGB2BGR))
cv2.waitKey(0)
 
# Visualize the pose segmentation mask.
# segmentation_mask = detection_result.segmentation_masks[0].numpy_view()
# visualized_mask = np.repeat(segmentation_mask[:, :, np.newaxis], 3, axis=2) * 255
# cv2.imshow('sean', visualized_mask)
# cv2.waitKey(0)

 

pose_landmarker_full.task

8.96MB

 

소스를 입력하고 실행한다.

 

원래 pose.jpg 이미지

 

MediaPipe로 처리한 이미지

 

 

0 - nose
1 - left eye (inner)
2 - left eye
3 - left eye (outer)
4 - right eye (inner)
5 - right eye
6 - right eye (outer)
7 - left ear
8 - right ear
9 - mouth (left)
10 - mouth (right)
11 - left shoulder
12 - right shoulder
13 - left elbow
14 - right elbow
15 - left wrist
16 - right wrist
17 - left pinky
18 - right pinky
19 - left index
20 - right index
21 - left thumb
22 - right thumb
23 - left hip
24 - right hip
25 - left knee
26 - right knee
27 - left ankle
28 - right ankle
29 - left heel
30 - right heel
31 - left foot index
32 - right foot index