Research map / 3D vision / Gaussian splatting
3D Gaussian Splatting Compression with Object Scalability
Ruixiang Xue, Tong Chen, Zhan Ma
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 overcomplete layer search (L_over=16 merged to L=8) for layer construction (ablation) | GoDe layered 3DGS · Mip-NeRF360 | BD-BR: -6.06% without finetuning | Improved |
| add progressive distillation top-down (L-1 to 1) (ablation) | RecastGS · Mip-NeRF360 | BD-BR: -31.27% vs finetuned GoDe | Improved |
| progressive distillation bottom-up instead of top-down (ablation) | RecastGS · Mip-NeRF360 | BD-BR: 21.22% vs 31.27% reduction | Hurt |
| use layered representation from overcomplete layer search for FCGS compression (ablation) | LayeredCGS · Mip-NeRF360 | BD-BR vs FCGS: -30.07% | Improved |
| add layer-wise context model (ablation) | LayeredCGS · Mip-NeRF360 | BD-BR: additional 8.06% reduction | Improved |
| replace G-PCC with RENO-GS for position coding (ablation) | LayeredCGS · Mip-NeRF360 | BD-BR and latency: additional 1.25% BD-BR; about 1.7x faster decode | Improved |
| region-adaptive allocation instead of uniform allocation | LayeredCGS · Mip-NeRF360 (bicycle, garden, kitchen) | ROI PSNR vs full PSNR: ROI BD-BR -28.57%/-51.96%/-33.75%; full-image degradation 4.59%/12.05%/7.06% | Mixed |
| apply progressive distillation then swap to sparsified GaussianSpa backbone | RecastGS-LayeredCGS · Mip-NeRF360/Tanks&Temples/Deep Blending | bitrate | Improved |
Closest papers by meaning
- ContextGS : Compact 3D Gaussian Splatting with Anchor Level Context Model (NeurIPS 2024 · similarity 0.92)
- Fast Feedforward 3D Gaussian Splatting Compression (ICLR 2025 · similarity 0.92)
- CAT-3DGS: A Context-Adaptive Triplane Approach to Rate-Distortion-Optimized 3DGS Compression (ICLR 2025 · similarity 0.92)
- Locality-aware Gaussian Compression for Fast and High-quality Rendering (ICLR 2025 · similarity 0.91)
- ZPressor: Bottleneck-Aware Compression for Scalable Feed-Forward 3DGS (NeurIPS 2025 · similarity 0.91)
- CodecSplat: Ultra-Compact Latent Coding for Feed-Forward 3D Gaussian Splatting (NeurIPS 2026 · similarity 0.91)
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.
