Research map / 3D vision / Gaussian splatting
MesonGS: Post-training Compression of 3D Gaussians via Efficient Attribute Transformation
Shuzhao Xie, Weixiang Zhang, Chen Tang, Yunpeng Bai, Rongwei Lu, Shjia Ge, Zhi Wang
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 finetuning after compression (ablation) | MesonGS · Synthetic-NeRF | PSNR (dB): 29.37 -> 31.75 | Improved |
| add finetuning after compression (ablation) | MesonGS · Mip-NeRF 360 | PSNR (dB): 27.20 -> 27.45 | Improved |
| replace rotation quaternions with Euler angles (ablation) | MesonGS · Synthetic-NeRF | PSNR (dB): 30.44 -> 30.44 | No clear effect |
| use Euler angles instead of covariance-based replacement (ablation) | MesonGS · Positive-definite covariance after quantization | fraction positive definite: ~50% -> 92% | Improved |
| use block quantization instead of per-channel quantization (ablation) | MesonGS · Mip-NeRF 360 (tau=50%) | PSNR (dB): 25.35 -> 27.20 | Improved |
| use block quantization instead of per-channel quantization (ablation) | MesonGS · Synthetic-NeRF (tau=66%) | PSNR (dB): 29.47 -> 29.60 | Improved |
| add view-dependent importance I_d to pruning score (ablation) | MesonGS pruning · Mip-NeRF 360 | PSNR: 27.01 -> 27.52 | Improved |
| add view-dependent importance I_d to pruning score (ablation) | MesonGS pruning · Synthetic-NeRF | PSNR: 30.39 -> 30.52 | Improved |
| apply RAHT to scale vectors at 8-bit quantization (ablation) | MesonGS · Mip-NeRF 360 / Synthetic-NeRF | rendering quality | Hurt |
| prune 33% instead of 66% of Gaussians at 8-bit quantization (ablation) | MesonGS · Mip-NeRF 360 (8-bit) | rendering quality | Hurt |
| prune Gaussians (66%) before compression (ablation) | MesonGS · Mip-NeRF 360 | size (MB): 641.73 -> 265.94 | Improved |
| add vector quantization of 1+D SH coefficients (ablation) | MesonGS · Mip-NeRF 360 | size (MB): 255.61 -> 79.35 | Improved |
Later papers that compared with this method
3 later accepted papers report MesonGS as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- RaPTGS: Render-Agnostic Post-Training Compression of 3D Gaussian Splatting (ECCV 2026 · better on 6 of 7 shared results)
- REFINE: Super-efficient Pruning for 3D Gaussian Splatting via Rendering-Free Primitive Importance (ECCV 2026 · better on 4 of 5 shared results)
- RAP: Fast Feedforward Rendering-Free Attribute-Guided Primitive Importance Score Prediction for Efficient 3D Gaussian Splatting Processing (CVPR 2026 · 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 0.92)
- Locality-aware Gaussian Compression for Fast and High-quality Rendering (ICLR 2025 · similarity 0.91)
- Fast Feedforward 3D Gaussian Splatting Compression (ICLR 2025 · similarity 0.91)
- Compressed 3D Gaussian Splatting for Accelerated Novel View Synthesis (CVPR 2024 · similarity 0.90)
- ContextGS : Compact 3D Gaussian Splatting with Anchor Level Context Model (NeurIPS 2024 · similarity 0.90)
- FlexGS: Train Once, Deploy Everywhere with Many-in-One Flexible 3D Gaussian Splatting (CVPR 2025 · similarity 0.90)
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.
