Research map / 3D vision / Depth and stereo
DiLiGenT-Pi: Photometric Stereo for Planar Surfaces with Rich Details - Benchmark Dataset and Beyond
Feishi Wang, Jieji Ren, Heng Guo, Mingjun Ren, Boxin Shi
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 |
|---|---|---|---|
| fine-tune uncalibrated SDPS-Net on near-planar PS RELIEF (ablation) | SDPS-Net · DiLiGenT-Pi | mean angular error (deg): 27.76 -> 16.03 | Improved |
| fine-tune uncalibrated SDPS-Net on near-planar PS RELIEF (ablation) | SDPS-Net · DiLiGenT | mean angular error (deg): 9.51 -> 17.80 | Hurt |
| fine-tune uncalibrated UPS-FCN on near-planar PS RELIEF (ablation) | UPS-FCN · DiLiGenT-Pi | mean angular error (deg): 21.54 -> 14.99 | Improved |
| fine-tune uncalibrated UPS-FCN on near-planar PS RELIEF (ablation) | UPS-FCN · DiLiGenT | mean angular error (deg): 15.37 -> 25.05 | Hurt |
| fine-tune calibrated PS-FCN on near-planar PS RELIEF (ablation) | PS-FCN · DiLiGenT-Pi | mean angular error (deg): 9.84 -> 8.80 | Improved |
| fine-tune calibrated PS-FCN on near-planar PS RELIEF (ablation) | PS-FCN · DiLiGenT | mean angular error (deg): 9.12 -> 11.75 | Hurt |
| use per-pixel network (CNN-PS) instead of all-pixel PS-FCN | calibrated learning-based photometric stereo · DiLiGenT-Pi | mean angular error (deg): 9.8 -> 9.2 | Improved |
| add attention-weighted detail loss to all-pixel PS-FCN (NormAttention-PSN) | PS-FCN · DiLiGenT-Pi | mean angular error (deg): 9.8 -> 9.2 | Improved |
| uncalibrated instead of calibrated light setting | learning-based photometric stereo · DiLiGenT-Pi | mean angular error (deg): 9.2 -> 21.5 | Hurt |
| use translucent or rough material instead of metallic for the same coin geometry (ablation) | CNN-PS and PS-FCN · same-shape reflectance ablation | mean angular error (deg) | Hurt |
| reject pixels outside intensity range (TH46) versus plain LSPS | LSPS · DiLiGenT-Pi | mean angular error (deg): 10.5 -> 11 | No clear effect |
Closest papers by meaning
- DiLiGenRT: A Photometric Stereo Dataset with Quantified Roughness and Translucency (CVPR 2024 · similarity 0.90)
- DiLiGenT102: A Photometric Stereo Benchmark Dataset With Controlled Shape and Material Variation (CVPR 2022 · similarity 0.88)
- Unbiased Photometric Stereo for Colored Surfaces: A Variational Approach (CVPR 2016 · similarity 0.87)
- Deep Learning Methods for Calibrated Photometric Stereo and Beyond (TPAMI 2024 · similarity 0.87)
- A Benchmark Dataset and Evaluation for Non-Lambertian and Uncalibrated Photometric Stereo (CVPR 2016 · similarity 0.87)
- 3D Gaussian Flats: Hybrid 2D/3D Photometric Scene Reconstruction (NeurIPS 2025 · similarity 0.85)
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.
