Research map / 3D vision / Gaussian splatting
StratoSplat: Taming Layered Regularities for Sparse Aerial 3D Gaussian Splatting
Zihan Gao, Lingling Li, Licheng Jiao, Fang Liu, Wenping Ma, Yuwei Guo, Shuyuan Yang
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 layered-memory guided initialization to vanilla 3DGS (ablation) | 3DGS · LEVIR-NVS (3 views) | PSNR: 20.20 -> 21.00 | Improved |
| use DUSt3R initialization instead of SfM (ablation) | 3DGS · LEVIR-NVS (3 views) | PSNR: 20.20 -> 18.65 | Hurt |
| use VGGT initialization instead of SfM (ablation) | 3DGS · LEVIR-NVS (3 views) | PSNR: 20.20 -> 16.43 | Hurt |
| use DA3, COLMAP MVS or Marigold init instead of SfM (ablation) | 3DGS · LEVIR-NVS (3 views) | PSNR: 20.20 -> 20.49 / 20.45 / 20.31 | No clear effect |
| add fixed multi-plane Gaussians on top of layered init (ablation) | StratoSplat · LEVIR-NVS (3 views) | PSNR: 21.00 -> 21.81 | Improved |
| use neural multi-plane Gaussians instead of fixed planes (ablation) | StratoSplat · LEVIR-NVS (3 views) | PSNR: 21.81 -> 23.64 | Improved |
| add depth-dependent opacity bias (ablation) | StratoSplat · LEVIR-NVS (3 views) | PSNR: 23.64 -> 24.48 | Improved |
| use layered memory instead of point cloud accumulation (ablation) | StratoSplat · 3DAS City Scene 0 | PSNR: 22.99 -> 23.21 | Improved |
| change view processing order (random vs coverage) (ablation) | StratoSplat · LEVIR-NVS Observation | PSNR: 21.10 -> 21.14 | No clear effect |
| add pose noise (L3) to all stages (ablation) | StratoSplat · LEVIR-NVS (3 views) | PSNR: 24.48 -> 19.11 | Hurt |
| use StratoSplat instead of FSGS | StratoSplat · LEVIR-NVS (3 views) | PSNR: 20.93 -> 24.48 | Improved |
| fine-tune feed-forward predictors, then optimize per scene | TranSplat · LEVIR-NVS (3 views) | novel-view PSNR: 16.79 -> 17.31 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- DropoutGS — DropoutGS: Dropping Out Gaussians for Better Sparse-view Rendering (CVPR 2025) · better on 1 of 1 shared result
- DNGaussian — DNGaussian: Optimizing Sparse-View 3D Gaussian Radiance Fields with Global-Local Depth Normalization (CVPR 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
- Geometry-Propagated Gaussian Splatting for Aerial Sparse Novel View Synthesis (ECCV 2026 · similarity 0.92)
- Horizon-GS: Unified 3D Gaussian Splatting for Large-Scale Aerial-to-Ground Scenes (CVPR 2025 · similarity 0.91)
- StructSplat: Generalizable 3D Gaussian Splatting from Uncalibrated Sparse Views (ECCV 2026 · similarity 0.89)
- GaussianSpa: An "Optimizing-Sparsifying" Simplification Framework for Compact and High-Quality 3D Gaussian Splatting (CVPR 2025 · similarity 0.89)
- SplatFormer: Point Transformer for Robust 3D Gaussian Splatting (ICLR 2025 · similarity 0.88)
- COLMAP-Free 3D Gaussian Splatting (CVPR 2024 · similarity 0.88)
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.
