Developed an object detection pipeline for sports analysis by benchmarking YOLOv5, YOLOv8, YOLOv9, and YOLOv11 on the SoccerNet dataset to accurately identify balls, players, and referees—facilitating comprehensive performance evaluation and data-driven tactical insights. Achieved the best results with YOLOv11, outperforming all other models with a Precision of 91.26%, Recall of 85.05%, mAP@50 of 89.6%, and mAP@50–95 of 64.89%.
