Reproduce “Towards Athlete Fatigue Assessment from Association Football Videos” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Xavier Bou, Nathan Correger, Alexandre Cloots, Cédric Gavage, Silvio Giancola, Cédric Schwartz, François Delvaux, Rudi Cloots and 2 more
Researchers from the University of Liège and collaborators explore the feasibility of assessing athlete fatigue in association football exclusively from monocular broadcast videos. They introduce a method to extract consistent kinematic signals (speed and acceleration) from vision-based player trajectories and demonstrate that derived acceleration-speed profiles can reflect fatigue-related performance shifts over match duration.
Key insights
- State-of-the-art vision-based Game State Reconstruction (GSR) can provide spatio-temporal signals of sufficient quality for fatigue-related analysis for reliably tracked athletes, achieving high correlations with ground truth.
- Proper post-processing, including temporal smoothing (Kalman filter) and windowed numerical differentiation over a sufficiently large temporal window (e.g., 1.64s), is crucial for mitigating noise and deriving consistent kinematic estimates from vision-based trajectories.
- A–S profiles derived from processed video data can qualitatively show shifts consistent with physiological fatigue over the duration of a match, suggesting their potential as non-intrusive fatigue indicators.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.