From Motion to Rally: Deep Learning for Badminton Video Analysis

This thesis proposed an automated system that segment badminton match videos into individual rallies. It detects when a serve starts and when a rally ends using player body movements and shuttlecock motion. The system uses artificial intelligence to analyze video frames without manual labeling and the results show that the proposed method can identify rally boundaries with high accuracy. This work reduces the time needed to analyze badminton matches and supports coaches and players in reviewing and analyzing performance in the match video.