RGS-SLAM: Robust Gaussian Splatting SLAM with One-Shot Dense Initialization

2025-12-28

arXiv.org

AI Summary

This work addresses the slow convergence and structural instability of conventional Gaussian splatting SLAM during early dense mapping, particularly in texture-rich or cluttered scenes, where residual-driven incremental Gaussian spawning proves inadequate. To overcome this limitation, we propose a training-free, single-pass dense initialization method that introduces, for the first time, a multi-view geometry–based strategy for one-shot Gaussian initialization. Our approach leverages DINOv3 features combined with a confidence-aware inlier classifier to establish dense multi-view correspondences, followed by triangulation to generate a structure-aware initial Gaussian distribution in a single step. This significantly enhances mapping stability and convergence speed, achieving state-of-the-art or comparable localization and reconstruction accuracy on the TUM RGB-D and Replica datasets, with higher rendering fidelity, approximately 20% faster convergence, and real-time performance maintained at up to 925 FPS.

Technology Category

Computer Vision: Multi-modal VisionSearch and Optimization: Learning to SearchMachine Learning: Multi-instance/Multi-view Learning

Application Category

Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for rankingGraph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphsSecurity and Privacy: Data transparency and provenance

Abstract

We introduce RGS-SLAM, a robust Gaussian-splatting SLAM framework that replaces the residual-driven densification stage of GS-SLAM with a training-free correspondence-to-Gaussian initialization. Instead of progressively adding Gaussians as residuals reveal missing geometry, RGS-SLAM performs a one-shot triangulation of dense multi-view correspondences derived from DINOv3 descriptors refined through a confidence-aware inlier classifier, generating a well-distributed and structure-aware Gaussian seed prior to optimization. This initialization stabilizes early mapping and accelerates convergence by roughly 20\%, yielding higher rendering fidelity in texture-rich and cluttered scenes while remaining fully compatible with existing GS-SLAM pipelines. Evaluated on the TUM RGB-D and Replica datasets, RGS-SLAM achieves competitive or superior localization and reconstruction accuracy compared with state-of-the-art Gaussian and point-based SLAM systems, sustaining real-time mapping performance at up to 925 FPS. Additional details and resources are available at this URL: https://breeze1124.github.io/rgs-slam-project-page/

Problem

• Gaussian splatting
• SLAM
• dense initialization
• 3D reconstruction
• real-time mapping

Innovation

• Gaussian Splatting
• One-Shot Initialization
• Dense Multi-View Correspondence
• Confidence-Aware Inlier Classification
• Real-Time SLAM