Overall framework
Abstract
SS4D is a native 4D generative model for synthesizing dynamic 3D objects from monocular video. It represents motion with structured spacetime latents, combines spatial consistency from image-to-3D priors with temporal reasoning layers, and compresses long latent sequences for efficient 4D generation.
Type
Publication
In ACM Transactions on Graphics (TOG)
Tong WU 吴桐
Assistant Professor @ Fudan
My research interests include 3d vision, long-tailed recognition, and robustness.