Kexin Tian

SUMO: Segment and Track Any Motion with Nonlinear State Space Models

Kexin Tian, Sixu Li, Keshu Wu, Yang Zhou, Zhengzhong Tu

Visual Object Tracking (VOT) and Moving Object Segmentation (MOS) are two fundamental tasks in computer vision that involve both spatial and temporal object dynamics. Existing methods rely predominantly on visual cues and thus often falter in real-world scenarios where object mot…Read more