Research map / 3D vision / Gaussian splatting
NanoGS: Training-Free and Lightweight Gaussian Splat Simplification
Butian Xiong, Rong Liu, Tiantian Zhou, Meida Chen, Zhiwen Fan, Andrew Feng
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 |
|---|---|---|---|
| use KNN-graph candidate selection instead of Octree-based selection (ablation) | NanoGS · Mip-NeRF 360 (rho=0.001) | PSNR (dB): 14.11 -> 17.20 | Improved |
| use KNN-graph candidate selection instead of Octree-based selection (ablation) | NanoGS · NeRF Synthetic (rho=0.001) | PSNR (dB): 15.85 -> 19.04 | Improved |
| add opacity-based filtering before merging (ablation) | NanoGS · Mip-NeRF 360 (rho=0.001) | PSNR (dB): 14.67 -> 17.20 | Improved |
| add opacity-based filtering before merging (ablation) | NanoGS · NeRF Synthetic (rho=0.001) | PSNR (dB): 18.87 -> 19.04 | Improved |
| add opacity-based filtering before merging (ablation) | NanoGS · NeRF Synthetic (rho=0.1) | PSNR (dB): 24.99 -> 25.81 | Improved |
| use I-divergence geometric cost instead of MSE surrogate (ablation) | NanoGS · Mip-NeRF 360 (rho=0.001) | PSNR (dB): 16.98 -> 17.20 | Improved |
| use I-divergence geometric cost instead of MSE surrogate (ablation) | NanoGS · NeRF Synthetic (rho=0.001) | PSNR (dB): 18.80 -> 19.04 | Improved |
| NanoGS vs best baseline (no fine-tuning), averaged over datasets | NanoGS · 4 benchmarks (rho=0.1) | PSNR (dB): +2.40 dB | Improved |
| NanoGS vs best baseline, rho=0.01 | NanoGS · 4 benchmarks (rho=0.01) | PSNR (dB): +4.84 dB | Improved |
| NanoGS vs best baseline, rho=0.001 | NanoGS · 4 benchmarks (rho=0.001) | PSNR (dB): +5.46 dB | Improved |
| increase Monte Carlo samples from S=1 to S=128 (ablation) | NanoGS · merge cost estimation | quality: S=1 vs S=128 | No clear effect |
| chain SOG compression on top of NanoGS compaction | NanoGS · 60 models storage | storage: 16.83 -> 2.45 (6.35x) | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- PUP3DGS — PUP 3D-GS: Principled Uncertainty Pruning for 3D Gaussian Splatting (CVPR 2025) · better on 4 of 4 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
- GaussianSpa: An "Optimizing-Sparsifying" Simplification Framework for Compact and High-Quality 3D Gaussian Splatting (CVPR 2025 · similarity 0.93)
- Speedy-Splat: Fast 3D Gaussian Splatting with Sparse Pixels and Sparse Primitives (CVPR 2025 · similarity 0.93)
- Optimized Minimal 3D Gaussian Splatting (NeurIPS 2025 · similarity 0.92)
- GaussianPro: 3D Gaussian Splatting with Progressive Propagation (ICML 2024 · similarity 0.92)
- PUP 3D-GS: Principled Uncertainty Pruning for 3D Gaussian Splatting (CVPR 2025 · similarity 0.92)
- IBGS: Image-Based Gaussian Splatting (NeurIPS 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.
