AI RESEARCH

Forecasting Motion in the Wild

arXiv CS.CV

ArXi:2604.01015v1 Announce Type: new Visual intelligence requires anticipating the future behavior of agents, yet vision systems lack a general representation for motion and behavior. We propose dense point trajectories as visual tokens for behavior, a structured mid-level representation that disentangles motion from appearance and generalizes across diverse non-rigid agents, such as animals in-the-wild.