How AI can track hockey action from faceoff to finish

University of Waterloo researchers have developed two AI systems that significantly improve the analysis of hockey games using video footage. The first system, Puck Localization Using Contextual Cues (PLUCC), enhances puck location tracking accuracy by 12% and reduces error by over 25%, providing a low-cost alternative to expensive tracking technologies. It works by analyzing players' gaze direction to infer puck location, offering accessibility for smaller teams and amateur organizations. The second system, SportMamba, improves multi-player tracking in sports videos, accounting for fast motion and obstructed views. Tested across multiple sports, it outperforms existing methods by 18%, enabling real-time performance analysis without the need for expensive equipment.