Research map / 3D vision / Gaussian splatting
EGGS: Explicitly Granular 3D Gaussian Splatting via Luma-Aware and Volume-Preserving Attribute Factorization
InGyu Jeong, Hyunmin Jung
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 explicit 1-dim luma DC instead of fully implicit DC (0-dim) (ablation) | EGGS · DC ablation (Fig. 7 table) | PSNR: 26.59 -> 27.02 | Improved |
| add explicit 1-dim luma DC instead of fully implicit DC (0-dim) (ablation) | EGGS · DC ablation (Fig. 7 table) | LPIPS: 0.242 -> 0.226 | Improved |
| store full 3-dim RGB DC instead of 1-dim luma (ablation) | EGGS · DC ablation (Fig. 7 table) | PSNR: 27.02 -> 27.10 | No clear effect |
| store full 3-dim RGB DC instead of 1-dim luma (ablation) | EGGS · DC ablation (Fig. 7 table) | storage size (MB): 4.62 -> 5.17 | Hurt |
| add volume-preserving scale factorization over naive 1D scale (ablation) | EGGS · Scale factorization ablation (Fig. 8 table) | SSIM: 0.827 -> 0.833 | Improved |
| add volume-preserving scale factorization over naive 1D scale (ablation) | EGGS · Scale factorization ablation (Fig. 8 table) | PSNR: 27.20 -> 27.18 | No clear effect |
| add volume-preserving scale factorization over naive 1D scale (ablation) | EGGS · Scale factorization ablation (Fig. 8 table) | storage size (MB): 4.86 -> 4.62 | Improved |
| add volume-preserving scale factorization over naive 1D scale (ablation) | EGGS · Train scene stability (10 trials) | success rate (%): 3 out of 10 (30%) -> 100% | Improved |
| add volume-preserving scale factorization over naive 1D scale (ablation) | EGGS · 13 scenes stability (10 trials each) | success rate (%): success rates dropping on 5 scenes -> 100% on all 13 scenes | Improved |
| increase local feature dimension D_L from 4 to 12 (ablation) | EGGS · Mip-NeRF360, T&T, Deep Blending RD curves | quality and file size | Mixed |
| reduce D_L from 12 to 8 in L configuration instead of switching to sparser configs (ablation) | EGGS-L · Mip-NeRF360, T&T, Deep Blending RD curves | SSIM at similar size | Improved |
| EGGS-XS versus OMG-XS on Mip-NeRF360 (smaller model) | EGGS-XS · Mip-NeRF360 | storage size (MB): 4.04 -> 2.67 | Improved |
Closest papers by meaning
- EGGS: Exchangeable 2D/3D Gaussian Splatting for Geometry-Appearance Balanced Novel View Synthesis (NeurIPS 2025 · similarity 0.91)
- Compressed 3D Gaussian Splatting for Accelerated Novel View Synthesis (CVPR 2024 · similarity 0.90)
- Optimized Minimal 3D Gaussian Splatting (NeurIPS 2025 · similarity 0.90)
- Locality-aware Gaussian Compression for Fast and High-quality Rendering (ICLR 2025 · similarity 0.90)
- Lightweight Predictive 3D Gaussian Splats (ICLR 2025 · similarity 0.89)
- Speedy-Splat: Fast 3D Gaussian Splatting with Sparse Pixels and Sparse Primitives (CVPR 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.
