Research map / 3D vision / Face and shape
DeepFaceFlow: In-the-Wild Dense 3D Facial Motion Estimation
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 DeepFaceFlow instead of LiteNet for 3D facial flow | 3D facial flow estimation · 4DFAB test | AEPE: 1.5 -> 1.3 | Helped |
| Use DeepFaceFlow instead of LiteNet for 3D facial flow | 3D facial flow estimation · Face3DVid test | AEPE: 2.2 -> 1.77 | Helped |
| Include all PNCC coordinate channels rather than only the z channel (ablation) | DeepFaceFlow · 4DFAB test | AEPE: 1.6 -> 1.3 | Helped |
| Include all PNCC coordinate channels rather than only the z channel (ablation) | DeepFaceFlow · Face3DVid test | AEPE: 1.971 -> 1.77 | Helped |
| Use DeepFaceFlow instead of the strongest scratch-trained baseline, LiteNet | 2D facial flow estimation · 4DFAB test | AEPE: 1.018 -> 0.941 | Helped |
| Use DeepFaceFlow instead of the strongest scratch-trained baseline, LiteNet | 2D facial flow estimation · Face3DVid test | AEPE: 1.268 -> 1.096 | Helped |
| Include all PNCC coordinate channels rather than only the z channel (ablation) | DeepFaceFlow · 4DFAB test | AEPE: 0.99 -> 0.941 | Helped |
| Include all PNCC coordinate channels rather than only the z channel (ablation) | DeepFaceFlow · Face3DVid test | AEPE: 1.176 -> 1.096 | Helped |
| Replace FlowNet2 with DeepFaceFlow 3D flow | Video-to-video self-reenactment · Self-reenactment videos 1–4 | average RGB distance: 7.5 -> 6.3; 9.5 -> 7.9; 8.7 -> 7.5; 9.2 -> 7.7 | Helped |
Closest papers by meaning
- 3D Face Tracking from 2D Video through Iterative Dense UV to Image Flow (CVPR 2024 · similarity 0.89)
- Depth-Aware Generative Adversarial Network for Talking Head Video Generation (CVPR 2022 · similarity 0.87)
- Flow-Guided One-Shot Talking Face Generation With a High-Resolution Audio-Visual Dataset (CVPR 2021 · similarity 0.87)
- 3D Human Tongue Reconstruction From Single "In-the-Wild" Images (CVPR 2022 · similarity 0.87)
- DEEPTalk: Dynamic Emotion Embedding for Probabilistic Speech-Driven 3D Face Animation (AAAI 2025 · similarity 0.86)
- OFER: Occluded Face Expression Reconstruction (CVPR 2025 · 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.