AnchorSplat: Feed-Forward 3D Gaussian Splatting with 3D Geometric Priors

TL;DR

AnchorSplat introduces a 3D scene reconstruction method using anchor-aligned Gaussian representations guided by geometric priors, achieving state-of-the-art results with fewer primitives and improved efficiency.

Contribution

It presents a novel anchor-aligned Gaussian framework guided by 3D priors, reducing complexity and enhancing scene reconstruction fidelity.

Findings

Outperforms previous methods on ScanNet++ v2 NVS benchmark.

Uses fewer Gaussian primitives for comparable or better quality.

Achieves more view-consistent reconstructions.

Abstract

Recent feed-forward Gaussian reconstruction models adopt a pixel-aligned formulation that maps each 2D pixel to a 3D Gaussian, entangling Gaussian representations tightly with the input images. In this paper, we propose AnchorSplat, a novel feed-forward 3DGS framework for scene-level reconstruction that represents the scene directly in 3D space. AnchorSplat introduces an anchor-aligned Gaussian representation guided by 3D geometric priors (e.g., sparse point clouds, voxels, or RGB-D point clouds), enabling a more geometry-aware renderable 3D Gaussians that is independent of image resolution and number of views. This design substantially reduces the number of required Gaussians, improving computational efficiency while enhancing reconstruction fidelity. Beyond the anchor-aligned design, we utilize a Gaussian Refiner to adjust the intermediate Gaussiansy via merely a few forward passes. Experiments on the ScanNet++ v2 NVS benchmark demonstrate the SOTA performance, outperforming previous methods with more view-consistent and substantially fewer Gaussian primitives.