Research map / 3D vision / Gaussian splatting
Drop-In Perceptual Optimization for 3D Gaussian Splatting
Ezgi Ozyilkan, Zhiqi Chen, Oren Rippel, Jona Ballé, Kedar Tatwawadi
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 |
|---|---|---|---|
| replace the original L1+SSIM loss with WD-R | 3DGS · Deep Blending (Indoors) | LPIPS: 0.243 -> 0.193 | Improved |
| add a lightly weighted original pixel loss to WD (WD-R) (ablation) | WD · Mip-NeRF 360 (Outdoors) | LPIPS: 0.228 -> 0.206 | Improved |
| add original-loss regularizer to WD to suppress web-like artifacts (ablation) | WD · Bicycle scene | visual artifacts | Improved |
| replace the loss with WD-R in Mip-Splatting | Mip-Splatting · Mip-NeRF 360 (Outdoors) | FID: 62.54 -> 45.87 | Improved |
| replace the loss with WD-R in Scaffold-GS | Scaffold-GS · Mip-NeRF 360 (Outdoors) | FID: 133.33 -> 67.74 | Improved |
| replace the loss with WD-R in Scaffold-GS | Scaffold-GS · Mip-NeRF 360 (Indoors) | CMMD: 0.391 -> 0.426 | Hurt |
| use composite loss instead of original loss | 3DGS · BungeeNeRF (Outdoors) | LPIPS: 0.098 -> 0.197 | Hurt |
| increase LPIPS weight in composite loss from 0.1 to 0.8 (ablation) | composite loss · Deep Blending (Indoors) | LPIPS: 0.235 -> 0.566 | Hurt |
| decrease WD pooling sigma from 4 to 1 (ablation) | WD · Mip-NeRF 360 (Indoors) | LPIPS: 0.152 -> 0.134 | Mixed |
| use saliency-guided adaptive sigma instead of constant sigma=4 (ablation) | WD · Mip-NeRF 360 (Indoors) | LPIPS: 0.152 -> 0.155 | No clear effect |
| replace the loss with WD (no regularizer) in Mip-Splatting | Mip-Splatting · BungeeNeRF | LPIPS: 0.113 -> 0.156 | Hurt |
| replace original loss with WD | 3DGS · Tanks & Temples | PSNR: 23.63 -> 22.53 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- Perceptual-GS — Perceptual-GS: Scene-adaptive Perceptual Densification for Gaussian Splatting (ICML 2025) · better on 4 of 5 shared results
- Mip-Splatting — Mip-Splatting: Alias-free 3D Gaussian Splatting (CVPR 2024) · better on 2 of 2 shared results
- Scaffold-GS — Scaffold-GS: Structured 3D Gaussians for View-Adaptive Rendering (CVPR 2024) · better on 2 of 2 shared results
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
- Perceptual-GS: Scene-adaptive Perceptual Densification for Gaussian Splatting (ICML 2025 · similarity 0.93)
- GaussianSpa: An "Optimizing-Sparsifying" Simplification Framework for Compact and High-Quality 3D Gaussian Splatting (CVPR 2025 · similarity 0.92)
- Quantifying and Alleviating Co-Adaptation in Sparse-View 3D Gaussian Splatting (NeurIPS 2025 · similarity 0.92)
- ReorgGS: Equivalent Distribution Reorganization for 3D Gaussian Splatting (NeurIPS 2026 · similarity 0.91)
- Temporal Smoothness-Aware Rate-Distortion Optimized 4D Gaussian Splatting (NeurIPS 2025 · similarity 0.91)
- 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.
