Research map / 3D vision / Gaussian splatting
Prune Wisely, Reconstruct Sharply: Compact 3D Gaussian Splatting via Adaptive Pruning and Difference-of-Gaussian Primitives
Haoran Wang, Guoxi Huang, Fan Zhang, David Bull, Nantheera Anantrasirichai
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 dynamic pruning ratio adjustment to refinement interval regulation (ablation) | 90%-pruned 3DGS · Mip-NeRF 360 | PSNR: 26.03 -> 26.17 | Improved |
| replace opacity-rank pruning score with Spatio-spectral Pruning Score (ablation) | 90%-pruned 3DGS (V2) · Mip-NeRF 360 | PSNR: 26.17 -> 26.99 | Improved |
| add 3D-DoG primitives after pruning (ablation) | 90%-pruned 3DGS (V3) · Mip-NeRF 360 | PSNR: 26.99 -> 27.16 | Improved |
| add 3D-DoG primitives after pruning (ablation) | 90%-pruned 3DGS (V3) · Mip-NeRF 360 | rendering FPS: 361.9 -> 289.0 | Hurt |
| add 3D-DoG primitives after pruning (ablation) | 90%-pruned 3DGS (V3) · Mip-NeRF 360 | training time: 13m28s -> 13m48s | Hurt |
| add refinement interval regulation with 90% pruning (ablation) | 3DGS · Mip-NeRF 360 | PSNR: 27.47 -> 26.03 | Hurt |
| add refinement interval regulation with 90% pruning (ablation) | 3DGS · Mip-NeRF 360 | training speed and FPS: 143.5 -> 362.4 FPS | Improved |
| apply full pruning method with 3D-DoG versus 3DGS | 3DGS · Deep Blending | PSNR: 29.75 -> 29.87 | Improved |
| use proposed method versus Speedy-Splat at similar model size | 90%-pruned 3DGS · Mip-NeRF 360 | PSNR: 26.84 -> 27.16 | Improved |
| use proposed method versus 3DGS | 3DGS · Tanks and Temples | training time: 17m50s -> 8m9s | Improved |
| use proposed method versus 3DGS | 3DGS · Mip-NeRF 360 | SSIM: 0.826 -> 0.789 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 4 on most shared results.
- Speedy-Splat — Speedy-Splat: Fast 3D Gaussian Splatting with Sparse Pixels and Sparse Primitives (CVPR 2025) · better on 5 of 5 shared results
- PuP-3DGS — PUP 3D-GS: Principled Uncertainty Pruning for 3D Gaussian Splatting (CVPR 2025) · better on 3 of 3 shared results
- GaussianSpa — GaussianSpa: An "Optimizing-Sparsifying" Simplification Framework for Compact and High-Quality 3D Gaussian Splatting (CVPR 2025) · better on 0 of 1 shared result
- MaskGaussian — MaskGaussian: Adaptive 3D Gaussian Representation from Probabilistic Masks (CVPR 2025) · better on 0 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
- Speedy-Splat: Fast 3D Gaussian Splatting with Sparse Pixels and Sparse Primitives (CVPR 2025 · similarity 0.94)
- PUP 3D-GS: Principled Uncertainty Pruning for 3D Gaussian Splatting (CVPR 2025 · similarity 0.94)
- 3D Gaussian Rendering Can Be Sparser: Efficient Rendering via Learned Fragment Pruning (NeurIPS 2024 · similarity 0.94)
- Optimized Minimal 3D Gaussian Splatting (NeurIPS 2025 · similarity 0.93)
- LP-3DGS: Learning to Prune 3D Gaussian Splatting (NeurIPS 2024 · similarity 0.93)
- CompactSplat: Spatially Adaptive Gaussian Distribution for Feedforward Scene Reconstruction (NeurIPS 2026 · 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.
