Research
Applied research in football analytics
Our R&D team conducts applied research in deep learning, computer vision and data analysis. We tackle the industry’s fundamental problems: from identifying players in a video stream to modelling game scenarios and detecting events automatically.
Research
Research areas
Player tracking and identification
Extracting player coordinates from video, reconstructing trajectories and keeping each player’s identity throughout the match.
Game event analysis
Automatic detection of passes, shots and tackles, linked to a specific player, time and position on the pitch.
Match structure reconstruction
Reconstructing a team’s playing structure: player positions, formation and how the team executes its game model on the pitch.
Research and results
5th place
A solution that reconstructs player coordinates from a broadcast and evaluates tracking quality.
Research and results
Score 64.38
Improved the solution after the challenge to a score of 64.38 — state of the art among published results.
Research and results
2nd place
Automatic detection of on-ball actions and their temporal localisation in a match.
Research and results
3rd place
At the international CVPR workshop with “LTPI: Long-term Player Identification from Single-Camera Football Video”
Research and results
2025
SoccerNet Game State Reconstruction
5th place
A solution that reconstructs player coordinates from a broadcast and evaluates tracking quality.
2026
SoccerNet Game State Reconstruction
Score 64.38
Improved the solution after the challenge to a score of 64.38 — state of the art among published results.
2026
SoccerNet Player-Centric Ball Action Spotting
2nd place
Automatic detection of on-ball actions and their temporal localisation in a match.
2026
CVPR Workshop
3rd place
At the international CVPR workshop with “LTPI: Long-term Player Identification from Single-Camera Football Video”