Real-Time Possessing Relationship Detection for Sports Analytics
Yinda Xu, Yonggang Peng · 2020
In this paper, we propose a novel algorithm for relationship detection. This task involves the tracking of a target object and human pose. The target object is tracked with a visual object tracker. The human poses are estimated via a keypoint detector while the person identities are preserved with a simple yet effective IoU tracker. Finally, a possessing relationship inference is made based on the position information of the tracked target and humans. This algorithm meets the real-time requirement by running at over 20 FPS and we give an application illustration in sports analytics.