LocusGS: Spatially Grounded Tokens for Feed-Forward 3D Gaussian Splatting

2026-08-13

AI Summary

This work addresses the poor alignment between generated Gaussians and scene geometry in existing query-based feed-forward 3D Gaussian splatting methods, which stems from insufficient spatial consistency. To resolve this, we introduce learnable 3D anchor states—comprising centers and support radii—into the query-based 3D Gaussian representation for the first time. These anchors guide query interaction, multi-view feature aggregation, and Gaussian generation, while a geometric bias from anchors to rays and local-region Gaussian decoding explicitly models spatial correlations. Under the same Gaussian budget, our method significantly improves novel view synthesis quality, and the learned anchors naturally form a coherent spatial layout, yielding Gaussians with enhanced structural consistency.

Abstract

Recent query-based feed-forward 3DGS methods represent a scene using learnable queries, each aggregating multi-view evidence and decoding a group of Gaussians. Ideally, different queries should specialize in coherent local regions of the scene. However, we observe that Gaussians decoded from the same query often scatter across distant scene regions, resulting in weak query-level spatial coherence and poor alignment with the scene structure. We attribute this behavior to the purely latent representation of existing Gaussian queries. To address this limitation, we introduce LocusGS, which augments each Gaussian query with a 3D anchor state consisting of a center and a support radius. The anchor state is progressively refined across decoder layers and is used throughout query interaction, multi-view feature aggregation, and Gaussian generation. Specifically, an anchor-to-ray geometric bias guides each query toward spatially relevant image observations, while anchor-centered decoding organizes its Gaussians within a local region. Experiments on novel view synthesis benchmarks show that LocusGS improves rendering quality over query-based Gaussian token baselines under the same Gaussian budget. Further analysis shows that the learned anchors form coherent spatial layouts and lead to more structured Gaussian distributions, demonstrating that explicit anchor states improve the spatial organization. Our project page: https://leo-frank.github.io/LocusGS_viewer.

Problem

• 3D Gaussian Splatting
• spatial coherence
• query-based representation
• scene structure
• Gaussian distribution

Innovation

• 3D Gaussian Splatting
• spatially grounded tokens
• anchor-based decoding
• feed-forward rendering
• novel view synthesis