Polycepta: Object-Centric Appearance Estimation for Multi-Object Tracking

Polycepta is a novel object-centric appearance state estimation framework for multi-object tracking (MOT) that reformulates appearance modeling as a recursive estimation problem, allowing for dynamic updates of appearance states for each tracked object. Operating at 90.57 Hz, it achieves a state-of-the-art MOTA of 92.27% on the KITTI benchmark when integrated into the RobMOT framework, demonstrating significant reductions in identity switches and improved tracking performance across datasets like KITTI, Waymo, and MOT17. This approach enhances robustness in MOT systems by enabling continuous refinement of appearance estimates, making it particularly valuable for practitioners dealing with real-time tracking challenges.