Research map / 3D vision / Face and shape
Lightweight Photometric Stereo for Facial Details Recovery
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 the two-stage face-prior and near-field-lighting method instead of UPS-FCN | Photometric stereo on Real Set with three images · Real Set | average angular error (degrees): 45.708 -> 6.579 | Helped |
| Use the two-stage face-prior and near-field-lighting method instead of SDPS-Net | Photometric stereo on Real Set with three images · Real Set | average angular error (degrees): 33.154 -> 6.579 | Helped |
| Use the two-stage face-prior and near-field-lighting method instead of UPS-FCN | Photometric stereo on Light Stage with three images · Light Stage | average angular error (degrees): 31.254 -> 5.007 | Helped |
| Use the two-stage face-prior and near-field-lighting method instead of SDPS-Net | Photometric stereo on Light Stage with three images · Light Stage | average angular error (degrees): 15.592 -> 5.007 | Helped |
| Use photometric stereo with near-point-light input rather than Extreme3D | Single-image 3D face reconstruction · Test set | geometric error: 1.77 -> 0.86 | Helped |
| Use photometric stereo with near-point-light input rather than DFDN | Single-image 3D face reconstruction · Test set | geometric error: 1.54 -> 0.86 | Helped |
| Add a proxy estimation network that provides facial geometry prior (ablation) | Normal estimation network · Test set, 1 input | average angular error (degrees): 14.843 -> 9.875 | Helped |
| Add a proxy estimation network that provides facial geometry prior (ablation) | Normal estimation network · Test set, 3 inputs | average angular error (degrees): 9.499 -> 6.154 | Helped |
| Add 5000 synthesized images through data augmentation to proxy training (ablation) | Proxy estimation network · Test set, 1 input | average angular error (degrees): 12.342 -> 9.875 | Helped |
| Add 5000 synthesized images through data augmentation to proxy training (ablation) | Proxy estimation network · Test set, 3 inputs | average angular error (degrees): 8.694 -> 6.154 | Helped |
| Use two input images rather than one | Proposed photometric stereo method · Test set | average angular error (degrees): 10.641 -> 8.245 | Helped |
| Use three input images rather than one | Proposed photometric stereo method · Test set | average angular error (degrees): 10.641 -> 6.498 | Helped |
Closest papers by meaning
- Sparse Photometric 3D Face Reconstruction Guided by Morphable Models (CVPR 2018 · similarity 0.90)
- Adaptive 3D Face Reconstruction From Unconstrained Photo Collections (CVPR 2016 · similarity 0.88)
- Deep Facial Non-Rigid Multi-View Stereo (CVPR 2020 · similarity 0.88)
- Detailed Facial Geometry Recovery from Multi-View Images by Learning an Implicit Function (AAAI 2022 · similarity 0.88)
- Learning Detailed Face Reconstruction From a Single Image (CVPR 2017 · similarity 0.88)
- Enhancing Face Recognition With Self-Supervised 3D Reconstruction (CVPR 2022 · 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.