Research map / 3D vision / Gaussian splatting
Pixel-GS Density Control with Pixel-aware Gradient for 3D Gaussian Splatting
Zheng Zhang, Wenbo Hu, Yixing Lao, Tong He, Hengshuang Zhao
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 |
|---|---|---|---|
| weight per-view gradient by covered pixel count (pixel-aware gradient) in the growth condition (ablation) | 3DGS* · Mip-NeRF 360 | LPIPS: 0.202 -> 0.176 | Improved |
| add pixel-aware gradient alone (ablation) | 3DGS* · Tanks & Temples | PSNR: 24.23 -> 21.80 | Hurt |
| add scaled gradient field (distance-based scaling) alone (ablation) | 3DGS* · Tanks & Temples | PSNR: 24.23 -> 24.34 | Improved |
| add scaled gradient field alone (ablation) | 3DGS* · Mip-NeRF 360 | LPIPS: 0.202 -> 0.202 | No clear effect |
| add scaled gradient field to pixel-aware gradient (ablation) | Pixel-GS · Tanks & Temples | PSNR: 21.80 -> 24.38 | Improved |
| lower growth threshold of 3DGS* from 2e-4 to 1.28e-4 instead of using pixel-aware gradient (ablation) | 3DGS* · Mip-NeRF 360 | LPIPS: 0.202 -> 0.181 (Pixel-GS 0.176) | Mixed |
| lower growth threshold of 3DGS* to 1e-4 (ablation) | 3DGS* · Tanks & Temples | PSNR: 24.19 -> 23.86 | Hurt |
| use Pixel-GS instead of 3DGS* | 3DGS* · Mip-NeRF 360 | FPS: 126 -> 89 | Hurt |
| drop up to 99% of initial SfM points | Pixel-GS vs 3DGS · Mip-NeRF 360 | LPIPS | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Mip-NeRF 360 — Mip-NeRF 360: Unbounded Anti-Aliased Neural Radiance Fields (CVPR 2022) · 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
- Steepest Descent Density Control for Compact 3D Gaussian Splatting (CVPR 2025 · similarity 0.94)
- Speedy-Splat: Fast 3D Gaussian Splatting with Sparse Pixels and Sparse Primitives (CVPR 2025 · similarity 0.92)
- GaussianPro: 3D Gaussian Splatting with Progressive Propagation (ICML 2024 · similarity 0.92)
- SplatFormer: Point Transformer for Robust 3D Gaussian Splatting (ICLR 2025 · similarity 0.92)
- Improving Gaussian Splatting with Localized Points Management (CVPR 2025 · similarity 0.92)
- Frequency-Aware Density Control via Reparameterization for High-Quality Rendering of 3D Gaussian Splatting (AAAI 2025 · similarity 0.92)
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.
