Research map / 3D vision / Gaussian splatting
Momentum-GS: Momentum Gaussian Self-Distillation for High-Quality Large Scene Reconstruction
Jixuan Fan, Wanhua Li, Yifei Han, Tianru Dai, Yansong Tang
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 parallel training with shared decoder (4 blocks) (ablation) | single-block baseline · Building | PSNR: 22.25 -> 23.10 | Improved |
| use shared decoder parallel training instead of independent training (4 blocks) (ablation) | block-wise training · Building | PSNR: 22.85 -> 23.10 | Improved |
| increase independent blocks from 4 to 8 (ablation) | independent training · Building | PSNR: 22.85 -> 23.23 | Improved |
| add scene momentum self-distillation (8 blocks) (ablation) | independent training with 8 blocks · Building | PSNR: 23.23 -> 23.56 | Improved |
| add reconstruction-guided block weighting (ablation) | Momentum-GS w/ self-distillation · Building | LPIPS: 0.205 -> 0.194 | Improved |
| alternate blocks on 4 GPUs instead of 8 GPUs parallel (ablation) | parallel training, 8 blocks · Building | PSNR: 23.34 -> 23.17 | Hurt |
| add momentum self-distillation to alternating training (ablation) | alternating 8 blocks on 4 GPUs · Building | PSNR: 23.17 -> 23.56 | Improved |
| weight blocks by PSNR+SSIM instead of PSNR only (ablation) | block weighting · Building | PSNR: 23.49 -> 23.65 | Improved |
| weight blocks by PSNR+SSIM instead of SSIM only (ablation) | block weighting · Building | LPIPS: 0.203 -> 0.194 | Improved |
| use momentum 0.9 instead of no momentum (0.0) (ablation) | teacher decoder · Building | PSNR: 23.44 -> 23.65 | Improved |
| raise momentum to 0.99 (ablation) | teacher decoder · Building | PSNR: 23.65 -> 22.06 | Hurt |
| vary lambda_consistency from 1 to 100 (ablation) | consistency loss weight · Building | PSNR: 23.53 -> 23.63 | No clear effect |
Later papers that compared with this method
3 later accepted papers report Momentum-GS as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- CAG-GS: Consistent Anchor Guided Gaussian Splatting for Large-scale Scene Rendering (AAAI 2026 · better on 5 of 11 shared results)
- DPGS: A Diffusion-Prior Guided Framework for Large-Scale 3D Gaussian Splatting Reconstruction (ECCV 2026 · better on 2 of 2 shared results)
- CityGS-X: A Scalable Architecture for Efficient and Geometrically Accurate Large-Scale Scene Reconstruction (ICCV 2025 · better on 3 of 5 shared results)
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
- Learning Efficient Fuse-and-Refine for Feed-Forward 3D Gaussian Splatting (NeurIPS 2025 · similarity 0.90)
- PUP 3D-GS: Principled Uncertainty Pruning for 3D Gaussian Splatting (CVPR 2025 · similarity 0.90)
- Speedy-Splat: Fast 3D Gaussian Splatting with Sparse Pixels and Sparse Primitives (CVPR 2025 · similarity 0.89)
- Scene Splatter: Momentum 3D Scene Generation from Single Image with Video Diffusion Model (CVPR 2025 · similarity 0.89)
- On Scaling Up 3D Gaussian Splatting Training (ICLR 2025 · similarity 0.89)
- MoDec-GS: Global-to-Local Motion Decomposition and Temporal Interval Adjustment for Compact Dynamic 3D Gaussian Splatting (CVPR 2025 · similarity 0.89)
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.
