Research map / 3D vision / Gaussian splatting
Plug-and-Play Optimization for 3D Gaussian Splatting Compression: Distribution Regularization, Probabilistic Pruning and Detail Compensation
Tian Bai, Zheng Qiu, Haojie Chen, Ziyang Dai
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 distribution regularization (variance + KL to standard normal) (ablation) | Scaffold-GS · Blender-Lego | PSNR: 35.51 -> 35.63 | Improved |
| add opacity-based probabilistic pruning (ablation) | Scaffold-GS · Blender-Lego | anchor count: 55K -> 37K | Improved |
| add opacity-based probabilistic pruning (ablation) | Scaffold-GS · DB-Drjohnson | PSNR: 29.69 -> 29.66 | Mixed |
| add high-frequency detail compensation network (ablation) | Scaffold-GS · Blender-Lego | PSNR: 35.51 -> 35.78 | Improved |
| add high-frequency detail compensation network (ablation) | Scaffold-GS · DB-Drjohnson | PSNR: 29.69 -> 29.81 | Improved |
| change anchor feature distribution (uniform, exponential, standard normal, custom) (ablation) | Scaffold-GS · Synthetic-NeRF lego | PSNR: 35.15 to 35.21 | No clear effect |
| apply plug-in modules on HAC at high rate | HAC highrate · Tanks&Temples | size (MB): 12.9 -> 9.5 | Improved |
| apply plug-in modules on HAC at high rate | HAC highrate · Synthetic-NeRF | PSNR: 33.72 -> 33.61 | Hurt |
| apply plug-in modules on HAC at low rate | HAC lowrate · Synthetic-NeRF | PSNR: 32.74 -> 32.54 | Hurt |
| apply plug-in modules on HAC | HAC · Tanks&Temples decoding time | decoding time (s): 15.05s -> 10.52s | Improved |
| use random probabilistic pruning instead of direct pruning (ablation) | Scaffold-GS · general | PSNR | 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.
- Scaffold-GS — Scaffold-GS: Structured 3D Gaussians for View-Adaptive Rendering (CVPR 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
- Fast Feedforward 3D Gaussian Splatting Compression (ICLR 2025 · similarity 0.94)
- Optimized Minimal 3D Gaussian Splatting (NeurIPS 2025 · similarity 0.94)
- Speedy-Splat: Fast 3D Gaussian Splatting with Sparse Pixels and Sparse Primitives (CVPR 2025 · similarity 0.93)
- Temporal Smoothness-Aware Rate-Distortion Optimized 4D Gaussian Splatting (NeurIPS 2025 · similarity 0.93)
- Compressed 3D Gaussian Splatting for Accelerated Novel View Synthesis (CVPR 2024 · similarity 0.93)
- 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.
