-
Developed a real-time sports analysis system for badminton using a hybrid approach combining YOLOv8 for player detection and DeepSORT for multi-frame tracking.
-
Built and annotated a custom dataset consisting of 200 clips from various real match footage to train the detection model, ensuring accurate performance across varied scenarios.
-
Extracted key player movement metrics such as speed, acceleration, distance, and zone coverage for performance evaluation.
-
Used LSTM for tactical zone prediction, achieving over 90% accuracy in anticipating player movement across court zones.
