Research map / 3D vision / Gaussian splatting
Predictive Photometric Uncertainty in Gaussian Splatting for Novel View Synthesis
Chamuditha Jayanga Ahangama Galappaththige, Thomas Gottwald, Peter Stehr, Edgar Heinert, Niko Suenderhauf, Dimity Miller, Matthias Rottmann
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 view-dependence via SH degree 1 (ablation) | uncertainty channel · Mip-NeRF360 UE (novel views) | AUSE DSSIM: 0.310 -> 0.245 | Improved |
| raise SH degree from 2 to 3 (ablation) | uncertainty channel · Mip-NeRF360 UE (novel views) | AUSE DSSIM: 0.218 -> 0.214 | Improved |
| raise SH degree from 1 to 3 (ablation) | uncertainty channel · Mip-NeRF360 UE (novel views) | Pearson DSSIM: 0.498 -> 0.547 | Improved |
| add Bayesian-inspired regularization in dense capture (ablation) | uncertainty channel · Mip-NeRF360 dense capture | AUSE | No clear effect |
| add Bayesian-inspired regularization with 4 training views (ablation) | uncertainty channel · Mip-NeRF360 sparse (4 views) | AUSE | Mixed |
| increase regularization weight to 0.16-0.32 for L1 error (ablation) | uncertainty channel · Mip-NeRF360 sparse (4 views) | AUSE L1 | Improved |
| increase regularization weight to 10.24 for DSSIM error (ablation) | uncertainty channel · Mip-NeRF360 sparse (4 views) | AUSE DSSIM | Improved |
| replace Var3DGS with ours | uncertainty estimation · Mip-NeRF360 UE (novel views) | AUSE DSSIM: 0.495 -> 0.214 | Improved |
| replace FisherRF with ours | uncertainty estimation · Mip-NeRF360 UE (novel views) | Pearson DSSIM: 0.009 -> 0.547 | Improved |
| use predicted uncertainty to guide view selection instead of FisherRF | active view selection · Active view selection, Mip-NeRF360 | PSNR: 20.266 -> 20.676 | Improved |
| attenuate change map with uncertainty | Feature Diff. · Pose-agnostic SCD, PASLCD | mIoU: 0.278 -> 0.359 | Improved |
| attenuate change map with uncertainty | MV3DCD-ZS · Pose-agnostic SCD, PASLCD | mIoU: 0.382 -> 0.439 | Improved |
Closest papers by meaning
- Variational Multi-scale Representation for Estimating Uncertainty in 3D Gaussian Splatting (NeurIPS 2024 · similarity 0.90)
- Structure Consistent Gaussian Splatting with Matching Prior for Few-shot Novel View Synthesis (NeurIPS 2024 · similarity 0.89)
- Generative Sparse-View Gaussian Splatting (CVPR 2025 · similarity 0.89)
- 4D Gaussian Splatting in the Wild with Uncertainty-Aware Regularization (NeurIPS 2024 · similarity 0.89)
- Perceptual-GS: Scene-adaptive Perceptual Densification for Gaussian Splatting (ICML 2025 · similarity 0.89)
- GPS-Gaussian: Generalizable Pixel-wise 3D Gaussian Splatting for Real-time Human Novel View Synthesis (CVPR 2024 · 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.
