From Fouls to Fair Play: Automated Video Analysis to Detect Illegal Player Contact
The client is the second-oldest labor union of the four North American professional sports leagues and provides formal representation to players for compensation negotiations and other agreements. Established in 1956, it offers resources and support to players to negotiate contracts and settle disputes with teams or the league. It also provides financial and legal resources to players and promotes player welfare and safety.
The union faced challenges in analyzing illegal contacts between players during organized team activities (OTA). The existing method involved manual verification of OTA videos, which was cumbersome and required game experts to evaluate every video for 4 to 5 hours.
Xoriant Solution: Key Contributions
Xoriant proposed a solution to develop a neural network for player detection, player tracking, contact detection, and contact classification. The team integrated Azure ML with Cosmo DB with the development of advanced algorithms to train and build computer vision models to analyze, detect, and classify the contact into ‘legal’ and ‘illegal’.
The union achieved a 90% automation in the analysis of the organized team activities video footage without the requirement of a game expert, saved thousands of dollars in manual monitoring, could allocate its resources better and explore the possibility of extending the types of analytics covered.