Sports Video Analysis
Overview
The field turns sports footage into structured information about players, movement, events, tactics, and performance, supporting coaching, officiating, and media. Common approaches combine object detection and tracking with pose estimation, camera calibration, and temporal models such as transformers or interaction graphs for action spotting and trajectory analysis. Benchmarks including SoccerNet and SportsMOT test tracking and event recognition, while newer video-language systems generate commentary and answer questions across longer or multiple camera views. Recognition is improving, but reliable identity tracking, precise temporal grounding, and rule-based or tactical reasoning remain difficult in crowded, long-form, and unfamiliar settings.
Latest papers
83 papers9 papers in the last four weeks, down 36% on the four weeks before. 0.1% of all new papers.