Dr. Jitendra musale Abhishek Khomane Rahul Dhope Kundalik Gavhane Abhishek Adsul

In this paper, we introduce a new methodology to improve basketball skills and fitness using techniques of computer vision. This work extends previous attempts in the literature on player tracking and in classifying ball movements in other sports like baseball. In this study, we are going to use YOLO-a Real time Object Detection Framework combined with Darknet, a Convolutional Neural Network to detect and classify basketball players and track the possession of the ball from videos. Training experts with ground truth data has helped us overcome diverse challenges regarding the angles of camera changes and occlusion of players with out-of-frame movements for exact tracking even in dynamic conditions. Traditional YOLO is being extended to incorporate contextual information from surrounding video frames, further enhancing its player detection and action recognition capabilities. The system supports basketball trainers and players with detailed analytics about individual performances, including shot tracking, ball handling, and teamwork dynamics. This framework provides a solid basis for optimizing training regimens, enhancing decision- making processes, and building better in-game strategies. We confirm the accuracy of our model and show the potential impact it could have on basketball performance analysis by comparing our data from experts.

https://doi.org/10.58357/jedcs.2025.v4.i1.6