Jinyuan Li ¹ and Jonathan Thamrun ² , ¹ Shenzhen College of International Education, No. 3 Antuoshan Sixth Road, Xiangmihu Street, Futian District, Shenzhen, Guangdong 518043, ² University of California, Irvine, Irvine, CA 92697

Badminton is one of the world’s most popular racket sports, yet accessible automated game analysis tools remain unavailable to recreational players. This paper presents MatchMotion, a cross-platform mobile application that enables users to upload badminton match videos for automated AI-powered analysis. The system employs three specialized YOLOv8 object detection models to identify players, shuttlecocks, and court boundaries within video frames, with a Python Flask backend hosted on AWS performing frame-by-frame inference and a Flutter frontend providing cross-platform mobile access. Custom algorithms analyze detection results to extract game statistics including rally length, hit counts, shuttle landing positions, and rally winners. Experimental evaluations on ten BWF World Championship rally videos yielded a mean shuttle detection accuracy of 37.6% and a mean hit recognition accuracy of 81.7%, demonstrating that velocity-based hit detection effectively compensates for frame-level detection limitations. MatchMotion democratizes badminton video analysis by providing automated insights previously accessible only through expensive professional systems.

Badminton video analysis, YOLOv8 object detection, Flutter mobile application, Shuttlecock tracking