LoopSplat: Loop Closure by Registering 3D Gaussian Splats

TL;DR

LoopSplat introduces a novel RGB-D SLAM method that uses 3D Gaussian Splats for dense mapping, enabling online loop closure and improved global consistency through direct submap registration and pose graph optimization.

Contribution

It is the first to perform online loop closure with 3D Gaussian Splats, enhancing accuracy and efficiency in dense 3D scene mapping.

Findings

Achieves superior tracking and mapping accuracy on multiple datasets.

Enables real-time loop closure with direct submap registration.

Demonstrates improved global consistency in dense 3D reconstructions.

Abstract

Simultaneous Localization and Mapping (SLAM) based on 3D Gaussian Splats (3DGS) has recently shown promise towards more accurate, dense 3D scene maps. However, existing 3DGS-based methods fail to address the global consistency of the scene via loop closure and/or global bundle adjustment. To this end, we propose LoopSplat, which takes RGB-D images as input and performs dense mapping with 3DGS submaps and frame-to-model tracking. LoopSplat triggers loop closure online and computes relative loop edge constraints between submaps directly via 3DGS registration, leading to improvements in efficiency and accuracy over traditional global-to-local point cloud registration. It uses a robust pose graph optimization formulation and rigidly aligns the submaps to achieve global consistency. Evaluation on the synthetic Replica and real-world TUM-RGBD, ScanNet, and ScanNet++ datasets demonstrates competitive or superior tracking, mapping, and rendering compared to existing methods for dense RGB-D SLAM. Code is available at loopsplat.github.io.