Research map / 3D vision / Gaussian splatting
KISS-GS: 3D Gaussian Splatting Compression Kept Simple
Wieland Morgenstern, Friedrich Elias Branschke, Florian Fleischmann, Adrian Szatmari, Paul Schlack, Florian Barthel, Anna Hilsmann, Peter Eisert
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 sparsification (GaussianSpa optimize-sparsify) to GaussianPOP pruning (ablation) | POPSpa compaction · Mip-NeRF 360 | PSNR: 27.69 -> 27.78 | Improved |
| add effective rank regularization to compaction (ablation) | POPSpa compaction · Mip-NeRF 360 | PSNR: 27.78 -> 27.79 | No clear effect |
| add SH AC k-means codebooks (ablation) | SOG-XT · Bicycle, 256k primitives | file size (MB): 7.313 -> 3.878 | Improved |
| add WebP image compression of attribute grids (ablation) | SOG-XT · Bicycle, 256k primitives | file size (MB): 6.908 -> 3.878 | Improved |
| add quaternion u8 (q=99) quantization (ablation) | SOG-XT · Bicycle, 256k primitives | file size (MB): 6.427 -> 3.878 | Improved |
| add PLAS sorting of primitives (ablation) | SOG-XT · Bicycle, 256k primitives | file size (MB): 4.493 -> 3.878 | Improved |
| add PLAS sorting of SH codebook centroids (ablation) | SOG-XT · Bicycle, 256k primitives | file size (MB): 3.920 -> 3.878 | Improved |
| add UV label packing for SH assignments (ablation) | SOG-XT · Bicycle, 256k primitives | file size (MB): 3.890 -> 3.878 | Improved |
| add PRAS covariance symmetry smoothing (ablation) | SOG-XT · Bicycle, 256k primitives | PSNR: -0.1104 PSNR without PRAS, size -11,810 bytes without PRAS | Improved |
| add signed-log remapping of means before 16-bit quantization (ablation) | SOG-XT · Bicycle, 256k primitives | PSNR: -7.5508 PSNR without it, size 3.359 MB vs 3.878 MB | Improved |
| add active mask with padding instead of pruning to square grid (ablation) | SOG-XT · Bicycle, 256k primitives | PSNR: -0.1150 PSNR when pruning to square, size +3,344 bytes | Improved |
| add PRAS so quaternions can use 100-level quantization instead of 256 levels (ablation) | SOG-XT · Bicycle, 256k primitives | file size (MB): 4.040 -> 3.878 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 4 on most shared results.
- Speedy-Splat (+10k) — Speedy-Splat: Fast 3D Gaussian Splatting with Sparse Pixels and Sparse Primitives (CVPR 2025) · better on 1 of 2 shared results
- GaussianSpa (40k) — GaussianSpa: An "Optimizing-Sparsifying" Simplification Framework for Compact and High-Quality 3D Gaussian Splatting (CVPR 2025) · better on 1 of 1 shared result
- PUP 3D-GS (+10k) — PUP 3D-GS: Principled Uncertainty Pruning for 3D Gaussian Splatting (CVPR 2025) · better on 1 of 1 shared result
- Mini-Splatting2 (28k) — Mini-Splatting: Representing Scenes with a Constrained Number of Gaussians (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
- Fast Feedforward 3D Gaussian Splatting Compression (ICLR 2025 · similarity 0.93)
- PUP 3D-GS: Principled Uncertainty Pruning for 3D Gaussian Splatting (CVPR 2025 · similarity 0.92)
- ContextGS : Compact 3D Gaussian Splatting with Anchor Level Context Model (NeurIPS 2024 · similarity 0.92)
- Optimized Minimal 3D Gaussian Splatting (NeurIPS 2025 · similarity 0.92)
- CodecSplat: Ultra-Compact Latent Coding for Feed-Forward 3D Gaussian Splatting (NeurIPS 2026 · similarity 0.92)
- Speedy-Splat: Fast 3D Gaussian Splatting with Sparse Pixels and Sparse Primitives (CVPR 2025 · 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.
