Injecting Inductive Bias to 3D Gaussian Splatting for Geometrically Accurate Radiance Fields
Keywords: 3D Gaussian Splatting, Surface Reconstruction
TL;DR: Geometrically Accurate 3D Gaussian Splatting
Abstract: 3D Gaussian Splatting (3DGS) has significantly advanced high-fidelity, real-time novel view synthesis. However, its discrete nature limits the accurate reconstruction of geometry. To address this issue, recent methods have introduced rendering and regularization of depth and normal maps from 3D Gaussians, leading to plausible results. In this paper, we argue that computing normals from independently trainable Gaussian covariances contradicts the strict definition of normals, which should instead be derived from the distribution of neighboring densities. To address this, we introduce an inductive bias into 3DGS by explicitly parameterizing covariances of Gaussians using principal axes and variances of distribution computed from neighboring Gaussians. These axes and variances are then regularized to ensure local surface smoothness. Our approach achieves competitive performance on multiple datasets.
Supplementary Material: zip
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: 1702
Loading