Research map / 3D vision / Depth and stereo
Uncalibrated Neural Inverse Rendering for Photometric Stereo of General Surfaces
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 image reconstruction with explicit interreflection modeling instead of ground-truth normal supervision | Ours · DiLiGenT average | MAE (degrees): 11.62; Chen et al. (2019): 9.51 | Mixed |
| Use image reconstruction with explicit interreflection modeling instead of ground-truth normal supervision | Ours · Proposed dataset average | MAE: 14.74; Nayar et al. (1991): 18.17 | Helped |
| Use estimated rather than ground-truth light directions and intensities | Ours · Proposed dataset average | MAE: 12.68 -> 14.74; +2.06 | Hurt |
| Replace classical initialization with robust partial-sum-of-singular-values initialization (ablation) | Ours · Cow | MAE | Helped |
| Remove weak-supervision loss during early optimization (ablation) | Ours · Cow | MAE | Hurt |
| Remove interreflection modeling and render using the initial normal map (ablation) | Ours · Vase | MAE | Hurt |
| Increase Gaussian image noise | Ours · Proposed dataset, noisy images | MAE | Hurt |
| Apply the method to a surface with deep concavities and large discontinuities | Ours · Rose | MAE: 60.82 degrees | Hurt |
Closest papers by meaning
- Neural Inverse Rendering for General Reflectance Photometric Stereo (ICML 2018 · similarity 0.94)
- DANI-Net: Uncalibrated Photometric Stereo by Differentiable Shadow Handling, Anisotropic Reflectance Modeling, and Neural Inverse Rendering (CVPR 2023 · similarity 0.91)
- Sparse Views Near Light: A Practical Paradigm for Uncalibrated Point-light Photometric Stereo (CVPR 2024 · similarity 0.88)
- RNb-NeuS: Reflectance and Normal-based Multi-View 3D Reconstruction (CVPR 2024 · similarity 0.88)
- MVCPS-NeuS: Multi-view Constrained Photometric Stereo for Neural Surface Reconstruction (CVPR 2024 · similarity 0.87)
- NeRFPrior: Learning Neural Radiance Field as a Prior for Indoor Scene Reconstruction (CVPR 2025 · 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.