CVPR Poster Uni4D: Unifying Visual Foundation Models for 4D

uni4d Uni4D is a novel method for learning discriminative 4D representations from point cloud videos using a self-disentangled Masked AutoEncoder. It outperforms existing methods. Uni4D is modular and any component can be swapped for other visual foundation model outputs. For custom vdeo depth estimation and dynamic masks, save them in

uni4d alternatif, Uni4D is modular and any component can be swapped for other visual foundation model outputs. For custom vdeo depth estimation and dynamic masks, save them in.