Research map / 3D vision / Gaussian splatting
VarSplat: Uncertainty-aware 3D Gaussian Splatting for Robust RGB-D SLAM
Anh Thuan Tran, Jana Kosecka
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add uncertainty weighting to tracking (ablation) | VarSplat · ScanNet (6 scenes) ablation | ATE RMSE (cm): 8.20 -> 7.63 | Improved |
| add uncertainty to loop detection (on top of tracking) (ablation) | VarSplat · ScanNet (6 scenes) ablation | ATE RMSE (cm): 7.63 -> 7.49 | Improved |
| add uncertainty to registration (on top of tracking+loop) (ablation) | VarSplat · ScanNet (6 scenes) ablation | ATE RMSE (cm): 7.49 -> 6.53 | Improved |
| freeze variance during tracking (ablation) | VarSplat · ScanNet (6 scenes) ablation | ATE RMSE (cm): 7.55 -> 6.53 | Improved |
| add depth residuals to variance loss (ablation) | VarSplat · ScanNet (6 scenes) ablation | ATE RMSE (cm): 7.17 -> 6.53 | Improved |
| use squared L2 instead of L1 for variance loss (ablation) | VarSplat · ScanNet (6 scenes) ablation | ATE RMSE (cm): 7.38 -> 6.53 | Improved |
| VarSplat instead of LoopSplat on ScanNet++ | VarSplat · ScanNet++ tracking | ATE RMSE (cm): 2.05 -> 1.69 | Improved |
| VarSplat instead of LoopSplat on Replica | VarSplat · Replica tracking | ATE RMSE (cm): 0.26 -> 0.23 | Improved |
| VarSplat instead of LoopSplat on Replica reconstruction | VarSplat · Replica reconstruction | Depth L1 / F1: F1 90.4 -> 90.2; Depth L1 0.51 -> 0.50 | No clear effect |
| VarSplat instead of Gaussian-SLAM on rendering | VarSplat · Replica rendering PSNR | PSNR: 42.08 -> 37.15 | Hurt |
| VarSplat instead of LoopSplat on ScanNet rendering | VarSplat · ScanNet rendering | PSNR: 24.92 -> 24.92 | No clear effect |
| add variance learning and rendering | VarSplat · runtime on Replica/Room0 | mapping time per frame (s): LoopSplat 1.9 vs 1.2 s mapping/frame | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- Loopy-SLAM — Loopy-SLAM: Dense Neural SLAM with Loop Closures (CVPR 2024) · better on 2 of 3 shared results
- GO-SLAM — GO-SLAM: Global Optimization for Consistent 3D Instant Reconstruction (ICCV 2023) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- DG-SLAM: Robust Dynamic Gaussian Splatting SLAM with Hybrid Pose Optimization (NeurIPS 2024 · similarity 0.91)
- Gaussian Splatting SLAM (CVPR 2024 · similarity 0.91)
- WildGS-SLAM: Monocular Gaussian Splatting SLAM in Dynamic Environments (CVPR 2025 · similarity 0.90)
- GS-SLAM: Dense Visual SLAM with 3D Gaussian Splatting (CVPR 2024 · similarity 0.90)
- VTGaussian-SLAM: RGBD SLAM for Large Scale Scenes with Splatting View-Tied 3D Gaussians (ICML 2025 · similarity 0.89)
- SplaTAM: Splat Track & Map 3D Gaussians for Dense RGB-D SLAM (CVPR 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
