Research map / 3D vision / Gaussian splatting
RAP: Fast Feedforward Rendering-Free Attribute-Guided Primitive Importance Score Prediction for Efficient 3D Gaussian Splatting Processing
Kaifa Yang, Qi Yang, Yiling Xu, Zhu Li
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 RAP scores instead of opacity thresholding for pruning | 3DGS · Mip-NeRF360 Outdoor | BD-Rate (%): -42.63 vs opacity baseline | Improved |
| use PUP-3DGS scores instead of opacity on Tanks&Temples | 3DGS · Tanks&Temples | BD-Rate (%): +7.06 | Hurt |
| use RAP scores in pruning-in-the-loop training | 3DGS · Mip-NeRF360 Outdoor | PSNR: 24.688 -> 24.709 | Improved |
| use RAP scores in pruning-in-the-loop training | 3DGS · Mip-NeRF360 Indoor | PSNR: 31.076 -> 30.696 | Hurt |
| pre-prune 20% of Gaussians with importance scores before MPEG GSC | GSC (G-PCC and video) · MPEG GSC R-D | BD-Rate: 15-20% BD-Rate gains for most methods | Improved |
| add opacity feature (ablation) | RAP · post-hoc pruning | PSNR: about 1-2 dB drop when removed | Improved |
| add Gaussian scale features (ablation) | RAP · post-hoc pruning | PSNR: about 0.5-1 dB drop when removed | Improved |
| add local or global normalization (each) (ablation) | RAP · post-hoc pruning | PSNR: 1.5-2 dB drop when either removed | Improved |
| add pruning-aware loss (ablation) | RAP · post-hoc pruning | PSNR: about 0.5 dB drop on average when removed | Improved |
| add distribution regularization loss (ablation) | RAP · post-hoc pruning, low retention | PSNR: approximately 0.25 dB | Improved |
| add color anisotropy and KNN distance features (ablation) | RAP · post-hoc pruning | PSNR | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 4 of 4 on most shared results.
- PUP-3DGS — PUP 3D-GS: Principled Uncertainty Pruning for 3D Gaussian Splatting (CVPR 2025) · better on 9 of 9 shared results
- LightGaussian — LightGaussian: Unbounded 3D Gaussian Compression with 15x Reduction and 200+ FPS (NeurIPS 2024) · better on 5 of 5 shared results
- EAGLES — EAGLES: Efficient Accelerated 3D Gaussians with Lightweight EncodingS (ECCV 2024) · better on 4 of 4 shared results
- MesonGS — MesonGS: Post-training Compression of 3D Gaussians via Efficient Attribute Transformation (ECCV 2024) · 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
- Optimized Minimal 3D Gaussian Splatting (NeurIPS 2025 · similarity 0.90)
- CodecSplat: Ultra-Compact Latent Coding for Feed-Forward 3D Gaussian Splatting (NeurIPS 2026 · similarity 0.90)
- LP-3DGS: Learning to Prune 3D Gaussian Splatting (NeurIPS 2024 · 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)
- Learning Efficient Fuse-and-Refine for Feed-Forward 3D Gaussian Splatting (NeurIPS 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.
