problem with fps

utils/record_dynamic.py

@@ -5,10 +5,11 @@ import sys
 import time
 from pathlib import Path
 import argparse
+import pandas as pd
 
 sys.path.append(str(Path(__file__).parent.parent))
 from skeleton.mediapipe_detector import MediaPipeDetector
-from ml_gestures.dynamic.feature_extractor import extract_sequence
+from ml_gestures_dynamic.feature_extractor import extract_sequence
 
 def main():
     parser = argparse.ArgumentParser()
@@ -17,6 +18,15 @@ def main():
     parser.add_argument('--camera', type=int, default=0)
     parser.add_argument('--duration', type=float, default=3.0, help='Длительность записи (сек)')
     args = parser.parse_args()
+    
+    output_path = Path(args.output)
+    
+    # Создаём файл с заголовком, если его ещё нет
+    if not output_path.exists():
+        with open(output_path, 'w', newline='') as f:
+            writer = csv.writer(f)
+            header = ['label', 'sequence_id', 'frame_idx'] + [f'f{i}' for i in range(99)]
+            writer.writerow(header)
 
     detector = MediaPipeDetector()
     cap = cv2.VideoCapture(args.camera)
@@ -26,8 +36,7 @@ def main():
 
     # Определяем следующий ID последовательности
     try:
-        import pandas as pd
-        df = pd.read_csv(args.output)
+        df = pd.read_csv(output_path)
         next_id = df['sequence_id'].max() + 1 if not df.empty else 0
     except:
         next_id = 0
@@ -46,27 +55,30 @@ def main():
         if result['success']:
             landmarks = result['landmarks']
             vis = detector.draw_landmarks(frame, result['pose_landmarks'])
-            if recording:
-                sequence.append(landmarks)
-                elapsed = time.time() - start_time
-                if elapsed >= args.duration:
-                    recording = False
-                    # Сохраняем последовательность
-                    seq_features = extract_sequence(sequence)
-                    with open(args.output, 'a', newline='') as f:
-                        writer = csv.writer(f)
-                        for i, feat in enumerate(seq_features):
-                            writer.writerow([args.label, next_id, i] + feat.tolist())
-                    print(f"Сохранено {len(sequence)} кадров для жеста {args.label}, ID={next_id}")
-                    sequence = []
-                    next_id += 1
         else:
             vis = frame
+            landmarks = None
 
+        # Если мы в режиме записи
         if recording:
+            if landmarks is not None:
+                sequence.append(landmarks)
+            elapsed = time.time() - start_time
+            # Отображаем прогресс записи
             cv2.putText(vis, f"RECORDING... {elapsed:.1f}/{args.duration}", (10, 30),
                         cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0,0,255), 2)
+            if elapsed >= args.duration:
+                recording = False
+                seq_features = extract_sequence(sequence)
+                with open(output_path, 'a', newline='') as f:
+                    writer = csv.writer(f)
+                    for i, feat in enumerate(seq_features):
+                        writer.writerow([args.label, next_id, i] + feat.tolist())
+                print(f"Сохранено {len(sequence)} кадров для жеста {args.label}, ID={next_id}")
+                sequence = []
+                next_id += 1
         else:
+            # Ожидание нажатия пробела
             cv2.putText(vis, f"Press SPACE to record '{args.label}'", (10, 30),
                         cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0,255,0), 2)
 
@@ -75,6 +87,36 @@ def main():
         if key == ord('q'):
             break
         if key == ord(' ') and not recording:
+            # Обратный отсчёт на живом видео
+            for i in range(3, 0, -1):
+                # Получаем свежий кадр для отсчёта
+                ret, frame = cap.read()
+                if not ret:
+                    break
+                frame = cv2.flip(frame, 1)
+                result = detector.detect(frame)
+                if result['success']:
+                    vis = detector.draw_landmarks(frame, result['pose_landmarks'])
+                else:
+                    vis = frame
+                cv2.putText(vis, f"Starting in {i}...", (frame.shape[1]//2-100, frame.shape[0]//2),
+                            cv2.FONT_HERSHEY_SIMPLEX, 1.5, (0,0,255), 3)
+                cv2.imshow('Record dynamic gesture', vis)
+                cv2.waitKey(1000)
+            # Показываем "GO!"
+            ret, frame = cap.read()
+            if ret:
+                frame = cv2.flip(frame, 1)
+                result = detector.detect(frame)
+                if result['success']:
+                    vis = detector.draw_landmarks(frame, result['pose_landmarks'])
+                else:
+                    vis = frame
+                cv2.putText(vis, "GO!", (frame.shape[1]//2-50, frame.shape[0]//2),
+                            cv2.FONT_HERSHEY_SIMPLEX, 2, (0,255,0), 3)
+                cv2.imshow('Record dynamic gesture', vis)
+                cv2.waitKey(500)
+            # Начинаем запись
             recording = True
             start_time = time.time()
             sequence = []