Spaces:
Paused
Paused
File size: 10,336 Bytes
bc18e51 a601b1d bc18e51 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 | """
Pose Analyzer - Core MediaPipe pose detection engine
Handles video frame processing and skeleton overlay generation
"""
import cv2
import numpy as np
import mediapipe as mp
from typing import List, Tuple, Optional, Dict, Any
from dataclasses import dataclass
import logging
from app.config import Config
from app.utils.helpers import timing_decorator
logger = logging.getLogger(__name__)
@dataclass
class PoseKeypoints:
"""Data class for storing pose keypoints from a single frame"""
landmarks: np.ndarray # Shape: (33, 3) - x, y, visibility
frame_number: int
timestamp: float
confidence: float # Average visibility score
class PoseAnalyzer:
"""
MediaPipe-based pose detection and analysis engine
Processes video frames to extract body keypoints and generate skeleton overlays
"""
# MediaPipe pose connections for skeleton drawing
POSE_CONNECTIONS = mp.solutions.pose.POSE_CONNECTIONS
def __init__(self, config: Optional[Dict[str, Any]] = None):
"""
Initialize pose analyzer with MediaPipe
Args:
config: Optional configuration dictionary (uses Config class defaults if None)
"""
self.config = config or Config.get_mediapipe_config()
# Initialize MediaPipe Pose
self.mp_pose = mp.solutions.pose
self.mp_drawing = mp.solutions.drawing_utils
self.mp_drawing_styles = mp.solutions.drawing_styles
# Create pose detector instance
self.pose = self.mp_pose.Pose(
static_image_mode=False,
model_complexity=self.config['model_complexity'],
smooth_landmarks=self.config['smooth_landmarks'],
min_detection_confidence=self.config['min_detection_confidence'],
min_tracking_confidence=self.config['min_tracking_confidence']
)
self.keypoints_history: List[PoseKeypoints] = []
logger.info("PoseAnalyzer initialized with MediaPipe Pose")
def process_frame(self, frame: np.ndarray, frame_number: int,
timestamp: float) -> Optional[PoseKeypoints]:
"""
Process a single video frame to detect pose
Args:
frame: BGR image from OpenCV
frame_number: Frame index in video
timestamp: Timestamp in seconds
Returns:
PoseKeypoints object if pose detected, None otherwise
"""
# Convert BGR to RGB for MediaPipe
rgb_frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
# Process frame with MediaPipe
results = self.pose.process(rgb_frame)
if results.pose_landmarks:
# Extract landmarks as numpy array
landmarks = self._extract_landmarks(results.pose_landmarks)
# Calculate average confidence (visibility score)
confidence = np.mean(landmarks[:, 2])
# Create PoseKeypoints object
pose_data = PoseKeypoints(
landmarks=landmarks,
frame_number=frame_number,
timestamp=timestamp,
confidence=confidence
)
return pose_data
return None
def _extract_landmarks(self, pose_landmarks) -> np.ndarray:
"""
Extract landmarks from MediaPipe results as numpy array
Args:
pose_landmarks: MediaPipe pose landmarks object
Returns:
Numpy array of shape (33, 3) containing x, y, visibility
"""
landmarks = []
for landmark in pose_landmarks.landmark:
landmarks.append([landmark.x, landmark.y, landmark.visibility])
return np.array(landmarks)
def draw_skeleton_overlay(self, frame: np.ndarray,
pose_keypoints: Optional[PoseKeypoints],
draw_confidence: bool = True) -> np.ndarray:
"""
Draw skeleton overlay on video frame
Args:
frame: Original BGR frame
pose_keypoints: Detected pose keypoints
draw_confidence: Whether to display confidence score
Returns:
Frame with skeleton overlay
"""
annotated_frame = frame.copy()
if pose_keypoints is None:
# Draw "No pose detected" message
cv2.putText(
annotated_frame,
"No pose detected",
(10, 30),
cv2.FONT_HERSHEY_SIMPLEX,
0.7,
(0, 0, 255),
2
)
return annotated_frame
# Only draw if confidence is above threshold
if pose_keypoints.confidence < Config.SKELETON_CONFIDENCE_THRESHOLD:
return annotated_frame
# Get frame dimensions
h, w = frame.shape[:2]
# Convert normalized coordinates to pixel coordinates
landmarks_px = pose_keypoints.landmarks.copy()
landmarks_px[:, 0] *= w # x coordinates
landmarks_px[:, 1] *= h # y coordinates
# Draw connections (skeleton lines)
for connection in self.POSE_CONNECTIONS:
start_idx, end_idx = connection
start_point = landmarks_px[start_idx]
end_point = landmarks_px[end_idx]
# Check visibility of both points
if (start_point[2] > Config.SKELETON_CONFIDENCE_THRESHOLD and
end_point[2] > Config.SKELETON_CONFIDENCE_THRESHOLD):
start_pos = (int(start_point[0]), int(start_point[1]))
end_pos = (int(end_point[0]), int(end_point[1]))
# Draw line with color gradient based on confidence
avg_confidence = (start_point[2] + end_point[2]) / 2
color = self._get_confidence_color(avg_confidence)
cv2.line(
annotated_frame,
start_pos,
end_pos,
color,
Config.SKELETON_LINE_THICKNESS,
cv2.LINE_AA
)
# Draw keypoints (circles)
for i, landmark in enumerate(landmarks_px):
if landmark[2] > Config.SKELETON_CONFIDENCE_THRESHOLD:
center = (int(landmark[0]), int(landmark[1]))
color = self._get_confidence_color(landmark[2])
cv2.circle(
annotated_frame,
center,
Config.SKELETON_CIRCLE_RADIUS,
color,
-1,
cv2.LINE_AA
)
# Draw confidence score
if draw_confidence:
confidence_text = f"Confidence: {pose_keypoints.confidence:.2f}"
cv2.putText(
annotated_frame,
confidence_text,
(10, 30),
cv2.FONT_HERSHEY_SIMPLEX,
0.7,
(0, 255, 0),
2
)
# Draw frame info
frame_text = f"Frame: {pose_keypoints.frame_number}"
cv2.putText(
annotated_frame,
frame_text,
(10, 60),
cv2.FONT_HERSHEY_SIMPLEX,
0.6,
(255, 255, 255),
1
)
return annotated_frame
def _get_confidence_color(self, confidence: float) -> Tuple[int, int, int]:
"""
Get color based on confidence score (green to yellow to red)
Args:
confidence: Confidence score (0-1)
Returns:
BGR color tuple
"""
if confidence >= 0.8:
return (0, 255, 0) # Green - high confidence
elif confidence >= 0.6:
return (0, 255, 255) # Yellow - medium confidence
else:
return (0, 165, 255) # Orange - low confidence
@timing_decorator
def process_video_batch(self, frames: List[np.ndarray],
start_frame_number: int,
fps: float) -> List[Optional[PoseKeypoints]]:
"""
Process a batch of video frames efficiently
Args:
frames: List of BGR frames
start_frame_number: Starting frame number
fps: Video frames per second
Returns:
List of PoseKeypoints (None for frames without detected pose)
"""
results = []
for i, frame in enumerate(frames):
frame_number = start_frame_number + i
timestamp = frame_number / fps
pose_data = self.process_frame(frame, frame_number, timestamp)
results.append(pose_data)
if pose_data:
self.keypoints_history.append(pose_data)
logger.info(f"Processed {len(frames)} frames, detected pose in "
f"{sum(1 for r in results if r is not None)} frames")
return results
def get_keypoints_array(self) -> np.ndarray:
"""
Get all detected keypoints as a numpy array
Returns:
Array of shape (N, 33, 3) where N is number of detected frames
"""
if not self.keypoints_history:
return np.array([])
return np.array([kp.landmarks for kp in self.keypoints_history])
def get_average_confidence(self) -> float:
"""
Calculate average confidence across all processed frames
Returns:
Average confidence score
"""
if not self.keypoints_history:
return 0.0
confidences = [kp.confidence for kp in self.keypoints_history]
return np.mean(confidences)
def reset(self):
"""Reset keypoints history for new video processing"""
self.keypoints_history.clear()
logger.info("PoseAnalyzer reset")
def __del__(self):
"""Cleanup MediaPipe resources"""
if hasattr(self, 'pose'):
self.pose.close() |