File size: 20,764 Bytes
c6abe34 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 | """
Personal Analysis Pipeline - Individual skill analysis with pose estimation.
This module provides personal training video analysis focused on a single player,
extracting skill metrics like shot form, dribbling patterns, and movement quality.
"""
import os
import sys
import math
from typing import Dict, Any, List, Tuple, Optional
# Add parent directory for template imports
sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
async def run_personal_analysis(video_path: str, options: Optional[Dict[str, Any]] = None, video_id: Optional[str] = None) -> Dict[str, Any]:
"""
Run personal analysis pipeline on a training video.
Focuses on a single primary subject and extracts:
- Pose keypoints and joint angles
- Shot form analysis and success rate
- Dribbling patterns
- Movement metrics (speed, distance)
Args:
video_path: Path to the video file
options: Optional configuration dict with keys:
- detections_stride: Frame stride for detections
- max_detections: Max detections to return
- detect_shots: Whether to run shot detection (default: True)
Returns:
Dictionary containing personal analysis results
"""
from utils import read_video
from app.config import get_settings
from shot_detector import ShotDetector
from trackers import BallTracker
from configs import PERSONAL_MODEL_PATH
from analysis.skill_diagnostic import SkillDiagnosticService
settings = get_settings()
# Read video frames
video_frames = read_video(video_path)
total_frames = len(video_frames)
if total_frames == 0:
return {
"error": "Could not read video frames",
"total_frames": 0,
}
# Get video FPS
fps = 30
try:
import cv2
cap = cv2.VideoCapture(video_path)
fps = cap.get(cv2.CAP_PROP_FPS) or 30
cap.release()
except:
pass
duration_seconds = total_frames / fps
# Initialize YOLO pose model
try:
from ultralytics import YOLO
pose_model = YOLO(settings.pose_model_path)
has_pose = True
except Exception as e:
print(f"Could not load pose model: {e}")
has_pose = False
# Track all detections and select primary subject
all_detections = []
player_stats = {} # track_id -> {frames: int, total_area: float, ball_interaction: int}
if has_pose:
# Run pose detection with tracking
batch_size = 20
for i in range(0, len(video_frames), batch_size):
batch = video_frames[i:i+batch_size]
results = pose_model.track(batch, conf=0.5, classes=[0], persist=True) # Class 0 = person
# Check for ball in these frames to find interaction
# (Heuristic: who is closest to the ball or hoop later)
for frame_offset, result in enumerate(results):
frame_idx = i + frame_offset
if result.boxes is not None and len(result.boxes) > 0:
for j, box in enumerate(result.boxes):
bbox = box.xyxy[0].tolist()
track_id = int(box.id[0]) if box.id is not None else -1
if track_id == -1: continue
# Calculate bbox area
area = (bbox[2] - bbox[0]) * (bbox[3] - bbox[1])
if track_id not in player_stats:
player_stats[track_id] = {'frames': 0, 'total_area': 0, 'interaction_score': 0}
player_stats[track_id]['frames'] += 1
player_stats[track_id]['total_area'] += area
# Get keypoints
keypoints = None
if result.keypoints is not None and j < len(result.keypoints):
kp = result.keypoints[j].xy[0].tolist()
keypoints = kp
# Heuristic: person moving their arms significantly or near the center
# tends to be the player.
if len(kp) > 10:
wrist_y = kp[10][1]
shoulder_y = kp[6][1]
if wrist_y < shoulder_y: # Arm raised
player_stats[track_id]['interaction_score'] += 1
all_detections.append({
"frame": frame_idx,
"track_id": track_id,
"bbox": bbox,
"keypoints": keypoints,
})
# Select primary subject (Prioritize interaction > presence > size)
if player_stats:
scores = {}
for tid, stats in player_stats.items():
presence = stats['frames'] / total_frames
avg_size = stats['total_area'] / stats['frames']
interaction = stats['interaction_score'] / stats['frames']
# Weighted score
scores[tid] = (interaction * 1.0) + (presence * 0.5) + (avg_size / 200000 * 0.3)
primary_player = max(scores, key=scores.get)
else:
primary_player = None
# Filter detections for primary player
primary_detections = [d for d in all_detections if d["track_id"] == primary_player]
# Build per-frame detections for overlays (optional, can be large)
detections_stride = 1
max_detections = 200_000
if options:
try:
detections_stride = int(options.get("detections_stride", detections_stride))
except Exception:
pass
try:
max_detections = int(options.get("max_detections", max_detections))
except Exception:
pass
detections_stride = max(1, min(30, detections_stride))
max_detections = max(1_000, max_detections)
detections: List[Dict[str, Any]] = []
for det in primary_detections:
frame_num = int(det.get("frame", 0))
if frame_num % detections_stride != 0:
continue
bbox = det.get("bbox")
if not bbox:
continue
detections.append({
"frame": frame_num,
"object_type": "player",
"track_id": int(det.get("track_id", 0) or 0),
"bbox": bbox,
"confidence": 1.0,
"keypoints": det.get("keypoints"),
"team_id": None,
"has_ball": False,
})
if len(detections) >= max_detections:
break
# Analyze pose data for skill metrics
shot_attempts = 0
form_scores = []
dribble_count = 0
positions = []
knee_angles = []
elbow_angles = []
for det in primary_detections:
kp = det.get("keypoints")
if kp and len(kp) >= 17:
# Extract key joint positions (COCO keypoint format)
# 0:nose, 5:left_shoulder, 6:right_shoulder, 7:left_elbow, 8:right_elbow
# 9:left_wrist, 10:right_wrist, 11:left_hip, 12:right_hip
# 13:left_knee, 14:right_knee, 15:left_ankle, 16:right_ankle
# Calculate knee angle (for shot form)
left_knee_angle = calculate_angle(kp[11], kp[13], kp[15]) # hip-knee-ankle
right_knee_angle = calculate_angle(kp[12], kp[14], kp[16])
if left_knee_angle:
knee_angles.append(left_knee_angle)
if right_knee_angle:
knee_angles.append(right_knee_angle)
# Calculate elbow angle (for shooting form)
left_elbow_angle = calculate_angle(kp[5], kp[7], kp[9]) # shoulder-elbow-wrist
right_elbow_angle = calculate_angle(kp[6], kp[8], kp[10])
if left_elbow_angle:
elbow_angles.append(left_elbow_angle)
if right_elbow_angle:
elbow_angles.append(right_elbow_angle)
# Track wrist position for dribble detection
left_wrist = kp[9] if len(kp) > 9 else None
right_wrist = kp[10] if len(kp) > 10 else None
# Store position (hip center) for movement tracking
if len(kp) > 12:
hip_center = [
(kp[11][0] + kp[12][0]) / 2,
(kp[11][1] + kp[12][1]) / 2
]
positions.append({
"frame": det["frame"],
"position": hip_center
})
# Detect shot attempts (arm raise events)
shot_attempts = detect_shot_attempts(primary_detections)
# Detect dribbles (rapid vertical wrist movements)
dribble_count = detect_dribbles(primary_detections)
# Calculate movement metrics
total_distance = 0
speeds = []
for i in range(1, len(positions)):
prev_pos = positions[i-1]["position"]
curr_pos = positions[i]["position"]
frame_diff = positions[i]["frame"] - positions[i-1]["frame"]
# Calculate pixel distance
dist = math.sqrt((curr_pos[0] - prev_pos[0])**2 + (curr_pos[1] - prev_pos[1])**2)
# Convert to approximate meters (assuming standard court proportions)
# This is a rough estimate - 1 pixel ≈ 0.01 meters for normalized view
dist_meters = dist * 0.01
total_distance += dist_meters
# Calculate speed (m/s)
if frame_diff > 0:
time_diff = frame_diff / fps
speed = dist_meters / time_diff
speeds.append(speed)
# Calculate averages
avg_speed = sum(speeds) / len(speeds) if speeds else 0
max_speed = max(speeds) if speeds else 0
# Convert to km/h
avg_speed_kmh = avg_speed * 3.6
max_speed_kmh = max_speed * 3.6
# Calculate form consistency (standard deviation of angles)
form_consistency = 100 - min(100, calculate_consistency(elbow_angles) * 2)
# Calculate averages
avg_knee_angle = sum(knee_angles) / len(knee_angles) if knee_angles else None
avg_elbow_angle = sum(elbow_angles) / len(elbow_angles) if elbow_angles else None
# Dribbles per minute
dribble_frequency = (dribble_count / duration_seconds) * 60 if duration_seconds > 0 else 0
# Acceleration events (significant speed changes)
acceleration_events = 0
for i in range(1, len(speeds)):
accel = abs(speeds[i] - speeds[i-1])
if accel > 2: # Threshold for significant acceleration
acceleration_events += 1
# Shot Success Detection
shot_stats = {
'total_attempts': shot_attempts,
'total_made': 0,
'total_missed': 0,
'overall_percentage': 0.0,
'by_type': {},
'shots': []
}
# Run shot detection if enabled (default: True)
detect_shots = options.get('detect_shots', True) if options else True
if detect_shots:
try:
# Check if ball detector model exists
model_path = PERSONAL_MODEL_PATH
if model_path is None or not os.path.exists(str(model_path)):
# Try relative path if absolute fails
if os.path.exists('models/nbl_v2_combined.pt'):
model_path = 'models/nbl_v2_combined.pt'
else:
model_path = None
if model_path is None:
print(f"Warning: Ball detector model not found, skipping shot detection")
else:
# Initialize ball tracker and shot detector
ball_tracker = BallTracker(model_path)
shot_detector = ShotDetector(
hoop_detection_model_path=model_path,
min_shot_arc_height=50,
hoop_proximity_threshold=100,
trajectory_window=30,
success_time_window=45
)
# Track ball
ball_tracks = ball_tracker.get_object_tracks(
video_frames,
read_from_stub=False
)
ball_tracks = ball_tracker.interpolate_ball_positions(ball_tracks)
# Detect hoop (if model available, otherwise use heuristics)
hoop_detections = shot_detector.detect_hoop_locations(
video_frames,
read_from_stub=False
)
# Detect and analyze shots
shots = shot_detector.detect_shots(
ball_tracks,
hoop_detections,
fps=fps
)
# Calculate shot statistics
shot_stats = shot_detector.calculate_shot_statistics(shots)
# Add ball and hoop detections to the main detections list for visualization
for f_idx, tracks in enumerate(ball_tracks):
for b_id, b_track in tracks.items():
if 'bbox' in b_track:
detections.append({
"frame": f_idx,
"object_type": "ball",
"track_id": b_id,
"bbox": b_track['bbox']
})
for f_idx, hoop in enumerate(hoop_detections):
if hoop and 'bbox' in hoop:
detections.append({
"frame": f_idx,
"object_type": "hoop",
"track_id": 0,
"bbox": hoop['bbox']
})
# --- Skill Diagnostic Logic Integration ---
try:
diagnostic_service = SkillDiagnosticService()
# Convert primary_detections to the format expected by the service
pose_tracks_formatted = [{} for _ in range(total_frames)]
for det in primary_detections:
f = det['frame']
tid = det['track_id']
if 0 <= f < total_frames:
pose_tracks_formatted[f][tid] = {'keypoints': det['keypoints']}
# Analyze each shot and attach feedback
coached_shots = []
for s in shots:
# Find matching entry angle from shot detector result
analysis = diagnostic_service.analyze_single_shot(s, pose_tracks_formatted)
coached_shots.append({
**s,
'biometrics': analysis['biometrics'],
'faults': analysis['faults'],
'feedback': analysis['feedback']
})
# Update shot_stats to include coached details
shot_stats['shots'] = coached_shots
except Exception as diag_err:
print(f"Skill Diagnostic failed: {diag_err}")
# -------------------------------------------
except Exception as e:
print(f"Shot detection failed: {e}")
# Keep default shot_stats
# Training load score (composite metric)
training_load = min(100, (
(dribble_count * 0.5) +
(shot_stats['total_attempts'] * 5) +
(total_distance * 2) +
(acceleration_events * 1)
))
return {
"total_frames": total_frames,
"duration_seconds": duration_seconds,
"primary_player_frames": len(primary_detections),
# Skill metrics
"shot_attempts": shot_stats['total_attempts'],
"shots_made": shot_stats['total_made'],
"shots_missed": shot_stats['total_missed'],
"shot_success_rate": shot_stats['overall_percentage'],
"shot_form_consistency": round(form_consistency, 1),
"shot_breakdown_by_type": shot_stats['by_type'],
"shot_details": shot_stats.get('shots', []),
"dribble_count": dribble_count,
"dribble_frequency_per_minute": round(dribble_frequency, 1),
# Movement metrics
"total_distance_meters": round(total_distance, 1),
"avg_speed_kmh": round(avg_speed_kmh, 1),
"max_speed_kmh": round(max_speed_kmh, 1),
"acceleration_events": acceleration_events,
# Pose analysis
"avg_knee_bend_angle": round(avg_knee_angle, 1) if avg_knee_angle else None,
"avg_elbow_angle_shooting": round(avg_elbow_angle, 1) if avg_elbow_angle else None,
# Training load
"training_load_score": round(training_load, 1),
"detections": detections,
}
def calculate_angle(p1: List[float], p2: List[float], p3: List[float]) -> Optional[float]:
"""
Calculate angle at p2 given three points.
Args:
p1, p2, p3: Points as [x, y] coordinates
Returns:
Angle in degrees at p2, or None if invalid
"""
if not all([p1, p2, p3]) or len(p1) < 2 or len(p2) < 2 or len(p3) < 2:
return None
# Check for valid coordinates (not 0,0)
if p1[0] == 0 and p1[1] == 0:
return None
if p2[0] == 0 and p2[1] == 0:
return None
if p3[0] == 0 and p3[1] == 0:
return None
try:
v1 = [p1[0] - p2[0], p1[1] - p2[1]]
v2 = [p3[0] - p2[0], p3[1] - p2[1]]
dot = v1[0] * v2[0] + v1[1] * v2[1]
mag1 = math.sqrt(v1[0]**2 + v1[1]**2)
mag2 = math.sqrt(v2[0]**2 + v2[1]**2)
if mag1 * mag2 == 0:
return None
cos_angle = dot / (mag1 * mag2)
cos_angle = max(-1, min(1, cos_angle)) # Clamp to valid range
angle = math.degrees(math.acos(cos_angle))
return angle
except:
return None
def calculate_consistency(values: List[float]) -> float:
"""Calculate standard deviation as a measure of consistency."""
if not values or len(values) < 2:
return 0
mean = sum(values) / len(values)
variance = sum((x - mean) ** 2 for x in values) / len(values)
return math.sqrt(variance)
def detect_shot_attempts(detections: List[Dict]) -> int:
"""
Detect shot attempts by analyzing arm raise patterns.
A shot attempt is detected when the wrist rises significantly above
the shoulder and then drops.
"""
shots = 0
arm_raised = False
for det in detections:
kp = det.get("keypoints")
if not kp or len(kp) < 11:
continue
# Check right arm (more common for right-handed shooters)
shoulder_y = kp[6][1] if len(kp) > 6 else 0
wrist_y = kp[10][1] if len(kp) > 10 else 0
# Check if wrist is significantly above shoulder (negative Y is up)
if shoulder_y > 0 and wrist_y > 0:
if wrist_y < shoulder_y - 50: # Wrist 50+ pixels above shoulder
if not arm_raised:
arm_raised = True
elif wrist_y > shoulder_y:
if arm_raised:
shots += 1
arm_raised = False
return shots
def detect_dribbles(detections: List[Dict]) -> int:
"""
Detect dribbles by analyzing rapid vertical wrist movements.
"""
dribbles = 0
prev_wrist_y = None
direction = None # 'up' or 'down'
for det in detections:
kp = det.get("keypoints")
if not kp or len(kp) < 11:
continue
# Track dominant hand wrist
wrist_y = kp[10][1] if len(kp) > 10 and kp[10][1] > 0 else None
if wrist_y is None or prev_wrist_y is None:
prev_wrist_y = wrist_y
continue
diff = wrist_y - prev_wrist_y
# Detect direction change (dribble = down then up motion)
if diff > 10: # Moving down
if direction == 'up':
dribbles += 1
direction = 'down'
elif diff < -10: # Moving up
direction = 'up'
prev_wrist_y = wrist_y
return dribbles
|