Real2SAM2Real: Generative 3D Caches as Complementary Context for Video Diffusion 事件

Real2SAM2Real: Generative 3D Caches as Complementary Context for Video Diffusion arXiv:2606.00299v1 Announce Type: new Abstract: While Video Diffusion Models (VDMs) excel at synthesizing high-fidelity videos, enabling precise camera and scene control remains challenging. Existing methods predominantly rely on implicit diffusion priors to generate unobserved regions, inevitably leading to structural collapse during high-dynamic movements or complex occlusions. To address this challenge, we propo