A Deep Learning Approach to Badminton Player Footwork Detection Based on YOLO Models: A Comparative Study

Badminton is a dynamic sport which requires rapid decision making and precise movements so the analysis of “Forehand front corner, Backhand front corner, Forehand side, Backhand side, Forehand back court corner, Backhand back court corner” these 6 corners are crucial for enhancing performance of badminton players' footwork. Traditional methods are followed by manual observation and video analysis, are often slow and complicated. Advancement in computer vision and deep learning are bringing profound solutions in the field of sports like badminton. This study uses YOLOv8 and YOLOv9 models to detect and analyze badminton players' footwork patterns by creating a custom dataset from videos footage. Extracted frames from the videos have been annotated, trained and tested both models. In this research the YOLOv8 model achieved a mean Average Precision (mAP) of 0.633, where YOLOv9's 0.605. It also demonstrated higher precision, recall, F1-scores, and faster inference speeds, with fewer misclassifications. Visual inspections confirmed YOLOv8 better accuracy under various conditions. The results indicate that YOLOv8 can significantly enhance badminton coaching by providing precise, real-time feedback on player footwork. This research creating a bridge between artificial intelligence and practical sports training, advancing both fields and offering promising insights for coaches and athletes aiming to improve their performance.