Research map / 3D vision / Depth and stereo
Deep Learning Methods for Calibrated Photometric Stereo and Beyond
Yakun Ju, Kin‐Man Lam, Wuyuan Xie, Huiyu Zhou, Junyu Dong, 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 |
|---|---|---|---|
| replace Lambertian least-squares baseline with deep calibrated PS-FCN | DiLiGenT 96 images · DiLiGenT (96 images) | average mean angular error (degrees): 15.39 -> 8.39 | Improved |
| replace baseline with PX-Net (per-pixel, global illumination) | DiLiGenT 96 images · DiLiGenT (96 images) | average mean angular error (degrees): 15.39 -> 6.33 | Improved |
| use sparse 10-image input instead of 96 for PS-FCN | PS-FCN · DiLiGenT | average mean angular error (degrees): 8.39 -> 10.51 | Hurt |
| use sparse 10-image input instead of 96 for MF-PSN | MF-PSN · DiLiGenT | average mean angular error (degrees): 7.27 -> 8.48 | Hurt |
| add observation normalization to PS-FCN | PS-FCN · DiLiGenT (96 images) | average mean angular error (degrees): 8.39 -> 7.39 | Improved |
| add multi-scale fusion (MF-PSN) over PS-FCN | PS-FCN · DiLiGenT (96 images) | average mean angular error (degrees): 8.39 -> 7.27 | Improved |
| add attention and adaptive loss over PS-FCN | PS-FCN · DiLiGenT (96 images) | average mean angular error (degrees): 8.39 -> 7.92 | Improved |
| self-supervised IRPS compared with supervised PS-FCN | IRPS · DiLiGenT (96 images) | average mean angular error (degrees): 8.83 vs 8.39 | Mixed |
Closest papers by meaning
- Deep Photometric Stereo for Non-Lambertian Surfaces (TPAMI 2020 · similarity 0.94)
- Self-Calibrating Deep Photometric Stereo Networks (CVPR 2019 · similarity 0.93)
- A CNN Based Approach for the Point-Light Photometric Stereo Problem (IJCV 2022 · similarity 0.93)
- Learning to Minify Photometric Stereo (CVPR 2019 · similarity 0.92)
- CNN-PS: CNN-based Photometric Stereo for General Non-Convex Surfaces (ECCV 2018 · similarity 0.92)
- PS-FCN: A Flexible Learning Framework for Photometric Stereo (ECCV 2018 · similarity 0.92)
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.
