Research map / 3D vision / Face and shape
Joint Face Detection and Facial Motion Retargeting for Multiple Faces
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 multi-scale features in place of a shared single-scale representation (ablation) | SFN · LFW | NME (%): 2.16 -> 1.91 | Helped |
| Use multi-scale parameter-specific branches (ablation) | SFN · Expression test set | Mean expression error: 0.305 -> 0.229 | Helped |
| Use multi-scale features for MFN (ablation) | MFN with multi-scale SFN-generated labels · AFW | AP: 98.2 -> 98.8 | Helped |
| Use multi-scale SFN to generate the 3DMM training labels (ablation) | SS-MFN · AFW | AP: 97.5 -> 98.2 | Helped |
| Jointly regress face boxes and 3DMM parameters instead of training for detection only (ablation) | MFN · AFW | AP: 92.1 -> 98.8 | Helped |
| Use MS-SFN rather than SS-SFN (ablation) | SFN · AFLW2000-3D | NME (%): 4.31 -> 3.89 | Helped |
| Use multi-scale SFN for tracking | SFN · 300VW Scenario 3 | AUC: +9% over the second best method | Helped |
| Use the multi-face model instead of the single-face model for expression prediction | SFN versus MS-MFN · Expression test set | Mean expression error: 0.229 -> 0.288 | Hurt |
| Train one joint network rather than a separate detector followed by per-face retargeting | MFN · Google Pixel 2 | runtime: 49ms -> 39ms for 1 face; 109ms -> 39ms for 5 faces; 184ms -> 39ms for 10 faces | Helped |
Closest papers by meaning
- MVF-Net: Multi-View 3D Face Morphable Model Regression (CVPR 2019 · similarity 0.88)
- Towards Accurate Facial Motion Retargeting with Identity-Consistent and Expression-Exclusive Constraints (AAAI 2022 · similarity 0.87)
- RetinaFace: Single-Shot Multi-Level Face Localisation in the Wild (CVPR 2020 · similarity 0.87)
- Regressing Robust and Discriminative 3D Morphable Models With a Very Deep Neural Network (CVPR 2017 · similarity 0.87)
- Unsupervised Training for 3D Morphable Model Regression (CVPR 2018 · similarity 0.87)
- Learning Dense Correspondence for NeRF-Based Face Reenactment (AAAI 2024 · similarity 0.85)
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.