Research map / 3D vision / Gaussian splatting
3DGS-LM: Faster Gaussian-Splatting Optimization with Levenberg-Marquardt
Lukas Höllein, Aljaž Božič, Michael Zollhöfer, Matthias Nießner
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 |
|---|---|---|---|
| replace ADAM with LM optimizer in second stage | 3DGS · MipNeRF360 | training time (s): 1271 -> 972 | Improved |
| replace ADAM with LM optimizer in second stage | 3DGS · Tanks&Temples | training time (s): 736 -> 663 | Improved |
| replace ADAM with LM optimizer in second stage | 3DGS · Deep Blending | training time (s): 1222 -> 951 | Improved |
| replace ADAM with LM optimizer | DISTWAR · MipNeRF360 | training time (s): 966 -> 764 | Improved |
| replace ADAM with LM optimizer | gsplat · Tanks&Temples | training time (s): 646 -> 414 | Improved |
| replace ADAM with LM optimizer | Taming-3DGS · Deep Blending | training time (s): 447 -> 347 | Improved |
| replace ADAM with LM optimizer | 3DGS · MipNeRF360 | PSNR: 27.40 -> 27.39 | No clear effect |
| use L1/SSIM objective instead of L2 loss (ablation) | 3DGS + Ours · MipNeRF360 GARDEN | SSIM: 0.857 -> 0.863 | Improved |
| use L1/SSIM objective instead of L2 loss (ablation) | 3DGS (ADAM) · MipNeRF360 GARDEN | SSIM: 0.854 -> 0.862 | Improved |
| use L1/SSIM objective instead of L2 loss (ablation) | 3DGS + Ours · MipNeRF360 GARDEN | PSNR: 27.48 -> 27.29 | Hurt |
| increase images per LM batch from 40 to 100 (ablation) | 3DGS + Ours · NeRF-Synthetic | PSNR: 33.51 -> 33.77 | Improved |
| increase images per LM batch from 40 to 100 (ablation) | 3DGS + Ours · NeRF-Synthetic | GPU memory (GB): 15.4 -> 32.5 | Hurt |
Closest papers by meaning
- On Scaling Up 3D Gaussian Splatting Training (ICLR 2025 · similarity 0.92)
- Speedy-Splat: Fast 3D Gaussian Splatting with Sparse Pixels and Sparse Primitives (CVPR 2025 · similarity 0.92)
- GaussianSpa: An "Optimizing-Sparsifying" Simplification Framework for Compact and High-Quality 3D Gaussian Splatting (CVPR 2025 · similarity 0.92)
- FlashGS: Efficient 3D Gaussian Splatting for Large-scale and High-resolution Rendering (CVPR 2025 · similarity 0.92)
- Steepest Descent Density Control for Compact 3D Gaussian Splatting (CVPR 2025 · similarity 0.92)
- DashGaussian: Optimizing 3D Gaussian Splatting in 200 Seconds (CVPR 2025 · similarity 0.92)
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.
