Research map / 3D vision / Depth and stereo
DANI-Net: Uncalibrated Photometric Stereo by Differentiable Shadow Handling, Anisotropic Reflectance Modeling, and Neural Inverse Rendering
Read the paperCheck 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 |
|---|---|---|---|
| Add differentiable shadow handling based on iteratively optimized depth and light (ablation) | DANI-Net · DiLiGenT | normal MAE: 6.54 -> 6.87 | Helped |
| Add ASG anisotropic reflectance modeling instead of isotropic ISG bases (ablation) | DANI-Net · DiLiGenT$10^2$ anisotropic group, Al, Cu, and Steel | normal MAE: 18.42 -> 15.59; 12.53 -> 10.44; 16.23 -> 13.73 | Helped |
| Use weighted interpolation rather than Sobel, cross, or triangle normal fitting (ablation) | DANI-Net · DiLiGenT | normal MAE: 6.69 -> 6.54; 6.63 -> 6.54; 6.93 -> 6.54 | Helped |
| Use the staged schedule that gradually drops smoothness terms (ablation) | DANI-Net · DiLiGenT | normal MAE: 6.76; 6.82; 6.64 -> 6.54 | Helped |
| Add silhouette loss (ablation) | DANI-Net · DiLiGenT | normal MAE: 7.06 -> 6.54 | Helped |
| Keep silhouette loss throughout all stages for predominantly non-occluding silhouettes (ablation) | DANI-Net · DiLiGenT$10^2$ Nut Pp | normal MAE: 1.16° | Hurt |
| Initialize light with GCNet (ablation) | DANI-Net · DiLiGenT$10^2$ | normal MAE: 0.77° decrease | Helped |
| Remove low-intensity pixels below the 25th percentile (ablation) | DANI-Net · DiLiGenT$10^2$ Nylon and Acrylic | normal MAE: 5.18°; 2.24° | Helped |
| Replace the dynamic differentiable shadow with alternative shadow handling (ablation) | DANI-Net · DiLiGenT | normal MAE: 0.52° | Hurt |
Closest papers by meaning
- Uncalibrated Neural Inverse Rendering for Photometric Stereo of General Surfaces (CVPR 2021 · similarity 0.91)
- A Benchmark Dataset and Evaluation for Non-Lambertian and Uncalibrated Photometric Stereo (CVPR 2016 · similarity 0.90)
- RNb-NeuS: Reflectance and Normal-based Multi-View 3D Reconstruction (CVPR 2024 · similarity 0.88)
- Sparse Views Near Light: A Practical Paradigm for Uncalibrated Point-light Photometric Stereo (CVPR 2024 · similarity 0.87)
- DiLiGenRT: A Photometric Stereo Dataset with Quantified Roughness and Translucency (CVPR 2024 · similarity 0.86)
- Spin-UP: Spin Light for Natural Light Uncalibrated Photometric Stereo (CVPR 2024 · similarity 0.86)
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.