GTA: Global Tracklet Association for Multi-Object Tracking in Sports

Multi-object tracking in sports scenarios has become one of the focal pointsin computer vision, experiencing significant advancements through theintegration of deep learning techniques. Despite these breakthroughs,challenges remain, such as accurately re-identifying players upon re-entry intothe scene and minimizing ID switches. In this paper, we propose anappearance-based global tracklet association algorithm designed to enhancetracking performance by splitting tracklets containing multiple identities andconnecting tracklets seemingly from the same identity. This method can serve asa plug-and-play refinement tool for any multi-object tracker to further boosttheir performance. The proposed method achieved a new state-of-the-artperformance on the SportsMOT dataset with HOTA score of 81.04%. Similarly, onthe SoccerNet dataset, our method enhanced multiple trackers' performance,consistently increasing the HOTA score from 79.41% to 83.11%. These significantand consistent improvements across different trackers and datasets underscoreour proposed method's potential impact on the application of sports playertracking. We open-source our project codebase athttps://github.com/sjc042/gta-link.git.