Research map / 3D vision / Face and shape
Semi-Supervised Monocular 3D Face Reconstruction With End-to-End Shape-Preserved Domain Transfer
Jingtan Piao, Chen Qian, Hongsheng Li
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 |
|---|---|---|---|
| train generator and shape estimator end-to-end instead of separately (ablation) | full framework · AFLW-LFPA | NME3d | Improved |
| add landmark consistency loss to the domain-transfer generator (ablation) | full framework · AFLW-LFPA | NME3d | Improved |
| add edge-aware loss for in-the-wild images (ablation) | full framework · AFLW-LFPA | NME3d | Improved |
| map real images to rendered style instead of training on synthetic images with random background only (ablation) | shape estimator · AFLW-LFPA | NME3d | Improved |
| use domain-transferred real-domain-style synthetic images with transferred ground truth instead of the proposed framework (ablation) | shape estimator · AFLW-LFPA | NME3d | Hurt |
| use semi-supervised synthetic+unlabeled training instead of fully supervised 300WLP (PRN) on extreme yaw 60-90 | face reconstruction · AFLW2000-3D yaw 60-90 | 2D NME (%): 4.61 -> 4.11 | Improved |
| use semi-supervised method instead of fully supervised PRN on mean over all yaw | face reconstruction · AFLW2000-3D mean | 2D NME (%): 3.62 -> 3.88 | Hurt |
Closest papers by meaning
- Self-Supervised Monocular 3D Face Reconstruction by Occlusion-Aware Multi-view Geometry Consistency (ECCV 2020 · similarity 0.89)
- FML: Face Model Learning From Videos (CVPR 2019 · similarity 0.89)
- End-To-End 3D Face Reconstruction With Deep Neural Networks (CVPR 2017 · similarity 0.89)
- Semi-Supervised Single-View 3D Reconstruction via Prototype Shape Priors (ECCV 2022 · similarity 0.89)
- A Simple, Fast and Highly-Accurate Algorithm to Recover 3D Shape from 2D Landmarks on a Single Image (TPAMI 2017 · similarity 0.88)
- Beyond 3DMM: Learning to Capture High-Fidelity 3D Face Shape (TPAMI 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.
