GaussVid: Sparse-View Gaussian Splatting enhanced with 3D-Aware Video Diffusion Priors

zhenjun_zhao · x · 2026-08-26

GaussVid introduces a 3D-aware video restoration framework to improve sparse-view 3D Gaussian Splatting (3DGS) reconstruction. To address artifacts in sparse views, the authors constructed a large-scale 3DGS video dataset and introduced a camera-conditioned geometric prior. By using the first and last frames as boundary anchors and encoding camera relationships, the method injects spatial structure into the generation pipeline, ensuring geometrically grounded restoration across viewpoints. Experiments show state-of-the-art fidelity (PSNR/SSIM) and improved multi-view consistency.

Original post →

More from Multimodal

Multimodal channel →