GA-GS: Generation-Assisted Gaussian Splatting for Static Scene Reconstruction

Yedong Shen, Shiqi Zhang, Sha Zhang, Yifan Duan, Xinran Zhang, Wenhao Yu, Lu Zhang, Jiajun Deng, Yanyong Zhang · 2026

Read GA-GS: Generation-Assisted Gaussian Splatting for Static Scene Reconstruction by Yedong Shen, Shiqi Zhang, Sha Zhang, Yifan Duan, Xinran Zhang, Wenhao Yu, Lu Zhang, Jiajun Deng, Yanyong Zhang (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-04331.txt.

Description

Shen and colleagues rebuild static 3D scenes from moving-object video. Current methods lean on background alone and fail where moving objects hide parts. Their GA-GS method segments and removes moving regions, then uses a diffusion model to paint in hidden areas as pseudo-ground-truth supervision. A learnable authenticity score tunes opacity during rendering, balancing real and generated regions. The team built Trajectory-Match, recorded by a fixed-path robot with and without moving objects. GA-GS leads rivals on both, above all under large occlusions.

Catalog record

CREATORYedong Shen, Shiqi Zhang, Sha Zhang, Yifan Duan, Xinran Zhang, Wenhao Yu, Lu Zhang, Jiajun Deng, Yanyong Zhang
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE36903
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.04331

Browse the full library · About the Tower of Babel