PrismTrack: Perspective-Aware Multi-Cue Association for Robust Multi-Object Tracking
Multi-Object Tracking (MOT) remains challenging due to object occlusion, complex motions, and detection unreliability in crowded scenarios. We propose an enhanced MOT framework integrating and optimizing state-of-the-art components, specifically Improved Detection Confidence Boost (IDCBoost) and Track-Perspective-Based...