Research map / 3D vision / Depth and stereo
SPLINE-Net: Sparse Photometric Stereo Through Lighting Interpolation and Normal Estimation Networks
Qian Zheng, Yiming Jia, Boxin Shi, Xudong Jiang, Ling-Yu Duan, Alex C. Kot
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 Lighting Interpolation Network (compare Nets w/o loss with CNN-PS) (ablation) | NE-Net regression from sparse maps · DiLiGenT | angular error: 18.50 -> 11.09 | Improved |
| add symmetric loss (ablation) | Nets w/o loss · DiLiGenT | angular error: 11.09 -> 10.43 | Improved |
| add asymmetric loss (ablation) | Nets with L^s · DiLiGenT | angular error: 10.43 -> 10.35 | Improved |
| use SPLINE-Net instead of CNN-PS | CNN-PS retrained with 10 lights · CyclesPS-Test L17 | angular error: 19.76 -> 13.64 | Improved |
| use SPLINE-Net instead of CNN-PS | CNN-PS retrained with 10 lights · CyclesPS-Test L305 | angular error: 21.82 -> 17.07 | Improved |
| use SPLINE-Net instead of IW12 | IW12 · CyclesPS-Test L17 sphere specular | angular error: 2.23 -> 3.82 | Hurt |
| use SPLINE-Net instead of IW12 | IW12 · DiLiGenT ball | angular error: 3.33 -> 4.96 | Hurt |
| use SPLINE-Net instead of IA14 | IA14 · DiLiGenT average | angular error: 19.04 -> 10.35 | Improved |
| add Gaussian noise to light directions, sigma 0 to 2 degrees | SPLINE-Net · DiLiGenT | angular error: 10.35 -> 11.21 | Hurt |
| add Gaussian noise to light directions, sigma 0 to 8 degrees | SPLINE-Net · DiLiGenT | angular error: 10.35 -> 15.94 | Hurt |
| more outlier pixels in sparse inputs | SPLINE-Net · DiLiGenT | angular error | Hurt |
Later papers that compared with this method
2 later accepted papers report SPLINE-Net as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- A CNN Based Approach for the Point-Light Photometric Stereo Problem (IJCV 2022 · better on 1 of 1 shared result)
- PX-NET: Simple and Efficient Pixel-Wise Training of Photometric Stereo Networks (ICCV 2021 · better on 2 of 2 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Universal Photometric Stereo Network Using Global Lighting Contexts (CVPR 2022 · similarity 0.90)
- What is Learned in Deep Uncalibrated Photometric Stereo? (ECCV 2020 · similarity 0.90)
- Uncalibrated Neural Inverse Rendering for Photometric Stereo of General Surfaces (CVPR 2021 · similarity 0.89)
- PX-NET: Simple and Efficient Pixel-Wise Training of Photometric Stereo Networks (ICCV 2021 · similarity 0.88)
- DeepPS2: Revisiting Photometric Stereo Using Two Differently Illuminated Images (ECCV 2022 · similarity 0.88)
- Self-Calibrating Photometric Stereo by Neural Inverse Rendering (ECCV 2022 · 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.
