Research map / 3D vision / Face and shape
MoFA: Model-Based Deep Convolutional Face Autoencoder for Unsupervised Monocular Reconstruction
Ayush Tewari, Michael Zollhofer, Hyeongwoo Kim, Pablo Garrido, Florian Bernard, Patrick Perez, Christian Theobalt
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 landmark surrogate loss (AlexNet) (ablation) | MoFA · real data validation | landmark error (px): 5.3 -> 3.7 | Improved |
| add landmark surrogate loss (VGG-Face) (ablation) | MoFA · real data validation | landmark error (px): 4.9 -> 3.4 | Improved |
| add landmark surrogate loss (AlexNet) (ablation) | MoFA · real data validation | photometric error: 0.075 -> 0.078 | No clear effect |
| add landmark surrogate loss (VGG-Face) (ablation) | MoFA · real data validation | photometric error: 0.073 -> 0.078 | No clear effect |
| add landmark surrogate loss (ablation) | MoFA · Facewarehouse (9 identities) | geometric error (mm): 1.9mm -> 1.7mm | Improved |
| add landmark surrogate loss (ablation) | MoFA · Facewarehouse (9 identities) | landmark error (px): 5.0px -> 3.2px | Improved |
| add landmark surrogate loss (ablation) | MoFA · Facewarehouse (9 identities) | photometric error: 0.065 -> 0.068 | Hurt |
| use VGG-Face encoder instead of AlexNet (ablation) | MoFA · real data validation | landmark error (px): 5.3 -> 4.9 | Improved |
| use VGG-Face encoder instead of AlexNet (ablation) | MoFA · real data validation | photometric error: 0.075 -> 0.073 | Improved |
| unsupervised training on synthetic data | MoFA (AlexNet) · synthetic ground truth | geometric error (mm): 21.6mm -> 4.5mm | Improved |
| unsupervised training on synthetic data | MoFA (AlexNet) · synthetic ground truth | landmark error (px): 31.6 -> 3.9 | Improved |
| use analytic model-based decoder instead of learned convolutional decoder (ablation) | MoFA · qualitative (Fig. 14) | image fidelity | Improved |
Closest papers by meaning
- High-Fidelity Monocular Face Reconstruction Based on an Unsupervised Model-Based Face Autoencoder (TPAMI 2018 · similarity 0.97)
- InverseFaceNet: Deep Monocular Inverse Face Rendering (CVPR 2018 · similarity 0.88)
- Generating 3D Faces using Convolutional Mesh Autoencoders (ECCV 2018 · similarity 0.88)
- Self-Supervised Multi-Level Face Model Learning for Monocular Reconstruction at Over 250 Hz (CVPR 2018 · similarity 0.88)
- CNN-Based Real-Time Dense Face Reconstruction with Inverse-Rendered Photo-Realistic Face Images (TPAMI 2018 · similarity 0.88)
- Learning Detailed Face Reconstruction From a Single Image (CVPR 2017 · 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.
