By treating motion as structured rigid-body transformations rather than independent point displacements, SM4RT achieves better 4D reconstruction while being more computationally efficient and physically grounded.
SM4RT is a new AI model that reconstructs 3D scenes and understands how objects move from video. Unlike methods that track individual points separately, SM4RT recognizes that real objects move as rigid bodies following physics laws.