STATERA: Hidden Mass Estimation via Zero-Shot Sim-to-Real Kinematics using Frozen Temporal Tubelets
STATERA introduces an architecture to estimate an object's hidden Center of Mass (CoM) from video by separating physical motion from visual appearance. It uses frozen temporal representations and a specialized mixer to isolate the directional phase of the offset CoM, successfully achieving a high Physics Capture Ratio in real-world transfer, outperforming standard spatial models.