SoccerNet Team

SoccerNet is a research initiative focused on developing and benchmarking computer vision and audio processing techniques for analyzing soccer video and audio data. Current research emphasizes tasks like player and ball tracking, action recognition (including fouls), game state reconstruction (mapping player positions), and generating descriptive captions from video and audio commentary. These efforts leverage deep learning models, including variations of YOLO for object detection and tracking, and gradient-boosted trees for match prediction, contributing to advancements in video understanding and sports analytics. The resulting datasets and benchmarks facilitate the development of more robust and accurate automated analysis tools for soccer, with applications ranging from performance analysis to highlight generation.