Ball motion improves soccer header detection in broadcast videos
BMASH: Ball-Motion-Aware Soccer Header Spotting
Computer Vision and Pattern Recognition
Summary
Spotting when a soccer player heads the ball in broadcast videos is hard because these moments are quick and subtle. The researchers tested a strong existing video action recognition model called Video Swin for this task. They then created BMASH, a new method that adds information about the soccer ball’s position and movement to help detect headers more accurately. Their tests showed BMASH did better than Video Swin alone, especially by combining player actions with ball motion.
What this means in practice
- •For sports analytics teams: Automatically identify header events in soccer game footage to support performance analysis and event statistics.
- •For player safety analysts: Detect header occurrences in broadcast videos to assess and potentially improve player safety measures related to head impacts.
Authors
Ahmed Endris Hasen, Muhammad Shahzad Khan, Nikolaos Passalis, Jenni Raitoharju
Abstract
Recent advances in computer vision have made broadcast sports videos increasingly useful for event analysis, performance assessment, and player-safety applications. In soccer, however, header spotting remains a challenging problem due to the subtle and short-lived nature of header events. This paper focuses on soccer header spotting: identifying moments in broadcast videos where the ball contacts a player's head. We first adapt and evaluate Video Swin as a strong action-recognition baseline for this task, and then introduce BMASH, a ball-motion-aware fusion framework that integrates detector-derived ball features. BMASH combines Video Swin action representations with ball-presence and motion features from frame-level soccer-ball detection, integrating player-action context with ball dynamics to distinguish headers from visually similar events. We evaluate BMASH using game-level splits with separate test matches and rotating validation folds, considering both centered-window classification and continuous full-video spotting. Results show that Video Swin provides a strong baseline for header spotting, while BMASH improves clip-level AP and ROC-AUC over the corresponding Video Swin baseline. In continuous full-video spotting, BMASH achieves a comparable event-level F1-performance with a different precision--recall trade-off.