360-InpaintR: Reference-Guided 3D Inpainting for Unbounded Scenes

Chung-Ho Wu, Yang-Jung Chen, Chin-Yang Lin, Yi-Chuan Huang, Bo-Hsu Ke, Ying-Huan Chen, Jie-Ying Lee, Chun-Wei Tuan Mu, Min-Hung Chen, Yen-Yu Lin, Yu-Lun Liu

Keywords: NeRF inpainting, 3D Gaussian splatting, Reference-based inpainting, Unbounded scene

Abstract:

This paper introduces 360-InpaintR, the first reference-based 360° inpainting method for 3D Gaussian Splatting (3DGS) scenes, particularly designed for unbounded environments. Our method leverages multi-view information and introduces an improved unseen mask generation technique to address the challenges of view consistency and geometric plausibility in 360° scenes. We effectively integrate reference-guided 3D inpainting with diffusion priors to ensure consistent results across diverse viewpoints. To facilitate research in this area, we present a new 360° inpainting dataset and capture protocol, enabling high-quality novel view synthesis and quantitative evaluations of modified scenes. Experimental results demonstrate that 360-InpaintR performs favorably against existing methods in both quantitative metrics and qualitative assessments, particularly in complex scenes with large view variations.

Primary Area: applications to computer vision, audio, language, and other modalities

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 3151