Research map / 3D vision / Depth and stereo
MERLiN: Single-Shot Material Estimation and Relighting for Photometric Stereo
Ashish Tiwari, Satoshi Ikehata, Shanmuganathan Raman
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 gated attention fusion of skip connections (ablation) | MERLiN · Synthetic test set | roughness MSE (x10^-2): 18.071 -> 15.277 | Improved |
| add feature sharing between relighting and material decoders (ablation) | MERLiN · Synthetic test set | roughness MSE (x10^-2): 18.071 -> 17.156 | Improved |
| add global illumination network (ablation) | MERLiN · Synthetic test set | roughness MSE (x10^-2): 9.681 -> 8.267 | Improved |
| train and input with global-illumination images instead of direct-only images (ablation) | MERLiN · Synthetic test set | roughness MSE (x10^-2): 10.113 -> 8.267 | Improved |
| use four separate material decoders instead of one (ablation) | MERLiN · Synthetic test set | albedo MSE (x10^-2): 3.787 vs 3.781 | No clear effect |
| render relit images through BRDF model instead of relighting decoder (ablation) | MERLiN · Synthetic test set | relighting SSIM: 0.819 -> 0.894 | Improved |
| use 32 relit images from MERLiN instead of single image for photometric stereo | Fast-NFPS · LUCES | mean angular error (degrees): 17.06 -> 15.80 | Improved |
| use real images instead of MERLiN relit images for photometric stereo | Fast-NFPS · LUCES | mean angular error (degrees): 15.80 -> 14.11 | Improved |
Closest papers by meaning
- Universal Photometric Stereo Network Using Global Lighting Contexts (CVPR 2022 · similarity 0.89)
- Self-Calibrating Photometric Stereo by Neural Inverse Rendering (ECCV 2022 · similarity 0.88)
- DeepPS2: Revisiting Photometric Stereo Using Two Differently Illuminated Images (ECCV 2022 · similarity 0.88)
- ReLi3D: Relightable Multi-view 3D Reconstruction with Disentangled Illumination (ICLR 2026 · similarity 0.88)
- PRM: Photometric Stereo based Large Reconstruction Model (ICCV 2025 · similarity 0.88)
- Generative Multiview Relighting for 3D Reconstruction under Extreme Illumination Variation (CVPR 2025 · 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.
