uni4dUni4D addresses key challenges in existing methods, effectively modeling long-term motion representations in a latent space while preserving geometric information in 4D space.Uni4D is a framework that uses multiple pretrained vision models to understand dynamic scenes from casual videos. It performs dynamic 3D reconstruction, camera pose