G3Splat: Geometrically Consistent Generalizable Gaussian Splatting

TL;DR

G3Splat introduces a method for 3D scene reconstruction that enforces geometric priors, leading to more accurate and consistent results in novel-view synthesis and pose estimation, especially in zero-shot scenarios.

Contribution

The paper presents G3Splat, a novel approach that incorporates geometric priors to improve the accuracy and consistency of 3D Gaussian splats in scene reconstruction.

Findings

Achieves state-of-the-art results on RE10K for geometry and pose estimation.

Demonstrates strong zero-shot generalization on ScanNet.

Outperforms prior methods in geometry recovery and pose estimation.

Abstract

3D Gaussians have recently emerged as an effective scene representation for real-time splatting and accurate novel-view synthesis, motivating several works to adapt multi-view structure prediction networks to regress per-pixel 3D Gaussians from images. However, most prior work extends these networks to predict additional Gaussian parameters -- orientation, scale, opacity, and appearance -- while relying almost exclusively on view-synthesis supervision. We show that a view-synthesis loss alone is insufficient to recover geometrically meaningful splats in this setting. We analyze and address the ambiguities of learning 3D Gaussian splats under self-supervision for pose-free generalizable splatting, and introduce G3Splat, which enforces geometric priors to obtain geometrically consistent 3D scene representations. Trained on RE10K, our approach achieves state-of-the-art performance in (i) geometrically consistent reconstruction, (ii) relative pose estimation, and (iii) novel-view synthesis. We further demonstrate strong zero-shot generalization on ScanNet, substantially outperforming prior work in both geometry recovery and relative pose estimation. Code and pretrained models are released on our project page (https://m80hz.github.io/g3splat/).