AI-based technology that captures match and player data from broadcast video has been announced by software pioneer Stats.
The new AutoStats system can potentially capture tracking data from any game - past or present - with x/y coordinates, differentiating players and the ball, and coordinating player movements around the court or playing field. Deep player and team performance data can be tracked from video, even of games before optical tracking solutions were available.
The company started installing optical player tracking cameras in the rafters of NBA in 2009. Since then, chief technology officer Helen Sun says tracking data -an integral part of game preparations for analysts, media and fans - had been limited to stadiums with optical tracking cameras.
The new AI-powered technology was used last season by Orlando Magic, and will be used for college scouting. A new OpenPose layer - under license from Carnegie Mellon University - adds layers of body-pose information.
Stats has used optical tracking data which goes back to a 1999 soccer match to teach computers new aspects of the game, and hopes to unlock more data and insights through robust AI.