Research map / 3D vision / Face and shape
Beyond 3DMM: Learning to Capture High-Fidelity 3D Face Shape
Xiangyu Zhu, Chang Yu, Di Huang, Zhen Lei, Hao Wang, Stan Z. 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 |
|---|---|---|---|
| replace camera view network with model view network (normalize pose) (ablation) | FGNet · FG3D-test | NME: 3.57 -> 3.49 | Improved |
| replace model view network with virtual multiview network (lossless) (ablation) | FGNet · FG3D-test | NME: 3.49 -> 3.32 | Improved |
| apply VMN refinement on top of initial 3DMM fitting | 3DDFA initial shape · FG3D-test | NME: 6.11 -> 3.32 | Improved |
| use 3 views instead of 1 frontal view (ablation) | VMN · FG3D-test | NME: 3.63 -> 3.40 | Improved |
| use 5 views (add pitch variations) instead of 3 (ablation) | VMN · FG3D-test | NME: 3.40 -> 3.32 | Improved |
| add inverse yaw views (7 views instead of 5) (ablation) | VMN · FG3D-test | NME: 3.32 -> 3.35 | No clear effect |
| single frontalized view versus original image input (ablation) | VMN · FG3D-test | NME: 3.60 -> 3.63 | Hurt |
| replace face profiling with pose augmentation with depth completion and inpainting (ablation) | FGNet training data · FG3D-test | NME: 3.51 -> 3.32 | Improved |
| add shape transformation augmentation (120k to 474k samples) (ablation) | FGNet training data · FG3D-test | NME: 3.32 -> 2.80 | Improved |
| add plaster sculpture descriptor loss to MSE (ablation) | FGNet · FG3D-test | NME: 2.80 -> 2.75 | Improved |
| replace MSE with visual-guided distance (VGD) loss (ablation) | FGNet · FG3D-test | NME: 2.80 -> 2.66 | Improved |
| replace MSE with VGD loss (ablation) | FGNet · FG3D-test | DACE: 1.50 -> 1.43 | Improved |
Closest papers by meaning
- Nonlinear 3D Face Morphable Model (CVPR 2018 · similarity 0.92)
- MVF-Net: Multi-View 3D Face Morphable Model Regression (CVPR 2019 · similarity 0.92)
- On Learning 3D Face Morphable Model from In-the-Wild Images (TPAMI 2019 · similarity 0.91)
- Regressing Robust and Discriminative 3D Morphable Models With a Very Deep Neural Network (CVPR 2017 · similarity 0.91)
- Disentangling Features in 3D Face Shapes for Joint Face Reconstruction and Recognition (CVPR 2018 · similarity 0.91)
- Inequality-Constrained 3D Morphable Face Model Fitting (TPAMI 2023 · similarity 0.91)
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.
