Research map / 3D vision / Depth and stereo
Universal Photometric Stereo Network Using Global Lighting Contexts
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use global lighting contexts rather than a specific physical lighting model | UniPS-GLC · PS-Wild test, HDRI | MAE (degrees): 14.5 vs 24.8 | Helped |
| Use global lighting contexts rather than a specific physical lighting model | UniPS-GLC · PS-Wild test, Dir.+HDRI | MAE (degrees): 13.8 vs 31.5 | Helped |
| Use SwinT instead of ViL as the image feature extractor (ablation) | UniPS-GLC · PS-Wild test, Dir.+HDRI | MAE (degrees): 24.0 -> 16.1 | Helped |
| Use SwinT instead of ResNet101 as the image feature extractor (ablation) | UniPS-GLC · PS-Wild test, HDRI | MAE (degrees): 45.8 -> 16.8 | Helped |
| Communicate features before multi-scale fusion (ablation) | UniPS-GLC · PS-Wild test, Directional | MAE (degrees): 19.7 -> 17.0 | Helped |
| Communicate features before multi-scale fusion (ablation) | UniPS-GLC · PS-Wild test, HDRI | MAE (degrees): 16.8 -> 14.5 | Helped |
| Communicate during feature extraction (ablation) | UniPS-GLC · PS-Wild test, HDRI | MAE (degrees): 16.8 -> 20.1 | Hurt |
| Communicate after multi-scale feature fusion (ablation) | UniPS-GLC · PS-Wild test, Dir.+HDRI | MAE (degrees): 16.1 -> 21.1 | Hurt |
| Increase canonical resolution from 128×128 to 256×256 (ablation) | UniPS-GLC · PS-Wild test, HDRI | MAE (degrees): 18.1 -> 16.8 | Helped |
| Replace spatially-varying contexts with spatially-uniform contexts (ablation) | UniPS-GLC · PS-Wild test, Directional | MAE (degrees): 19.7 -> 40.5 (trival) | Hurt |
| Replace spatially-varying contexts with spatially-uniform contexts (ablation) | UniPS-GLC · PS-Wild test, HDRI | MAE (degrees): 16.8 -> 40.5 (trivial) | Hurt |
| Use three Transformer layers with PMA instead of max pooling (ablation) | UniPS-GLC · PS-Wild test, HDRI | MAE (degrees): 39.3 -> 16.8 | Helped |
Closest papers by meaning
- Scalable, Detailed and Mask-Free Universal Photometric Stereo (CVPR 2023 · similarity 0.93)
- A Benchmark Dataset and Evaluation for Non-Lambertian and Uncalibrated Photometric Stereo (CVPR 2016 · similarity 0.92)
- Light of Normals: Unified Feature Representation for Universal Photometric Stereo (ICLR 2026 · similarity 0.92)
- GPS-Net: Graph-based Photometric Stereo Network (NeurIPS 2020 · similarity 0.90)
- Sparse Views Near Light: A Practical Paradigm for Uncalibrated Point-light Photometric Stereo (CVPR 2024 · similarity 0.88)
- DiLiGenRT: A Photometric Stereo Dataset with Quantified Roughness and Translucency (CVPR 2024 · similarity 0.87)
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.