Personalize Your Gaussian: Consistent 3D Scene Personalization from a Single Image

Yuxuan Wang 1 Xuanyu Yi 1 Qingshan Xu 1 Yuan Zhou 1 Long Chen 2 Hanwang Zhang 1 1 Nanyang Technological University 2 Hong Kong University of Science and Technology · 2025

Read Personalize Your Gaussian: Consistent 3D Scene Personalization from a Single Image by Yuxuan Wang 1 Xuanyu Yi 1 Qingshan Xu 1 Yuan Zhou 1 Long Chen 2 Hanwang Zhang 1 1 Nanyang Technological University 2 Hong Kong University of Science and Technology (2025) 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-2505-14537.txt.

Description

Yuxuan Wang, Xuanyu Yi, Qingshan Xu, Yuan Zhou, Long Chen, and Hanwang Zhang personalize 3D scenes from a single photo. Their paper from twenty twenty-five works with 3D Gaussian Splatting. One photo shows just one view, so edits tend to warp as the camera moves. Past methods lack a way to spread the reference look to unseen angles.

Their CP-GS frame grows that look step by step. An image-to-3D generator first stretches the reference to new views. LoRA fine-tuning rounds then pull the model toward the reference style. Geometry cues keep the guide images in line across views. The final scene comes from this view-consistent guidance. Tests on real scenes show the bias largely gone. Quality beat prior methods by clear margins. The tool suits user-led edits where one photo must carry a whole scene.

Catalog record

CREATORYuxuan Wang 1 Xuanyu Yi 1 Qingshan Xu 1 Yuan Zhou 1 Long Chen 2 Hanwang Zhang 1 1 Nanyang Technological University 2 Hong Kong University of Science and Technology
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE46 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2505-14537

Browse the full library · About the Tower of Babel