Research map / 3D vision / Face and shape
Face2Faceρ: Real-Time High-Resolution One-Shot Face Reenactment
Kewei Yang, Kang Chen, Daoliang Guo, Song-Hai Zhang, Yuan-Chen Guo, Weidong Zhang
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 skip connections in u-shaped rendering network (ablation) | Face2Faceρ · Reenactment | FID: 72.5 -> 44.5 | Improved |
| add head pose injection into encoder and decoder (ablation) | Face2Faceρ · Reenactment | ARD: 3.83 -> 2.82 | Improved |
| use 3DMM head pose encoding instead of 2D landmarks (ablation) | Face2Faceρ · Reenactment | FID: 50.9 -> 44.5 | Improved |
| use 3DMM head pose encoding instead of 3D landmarks (ablation) | Face2Faceρ · Reenactment | FID: 43.4 vs 44.5 | No clear effect |
| use 3DMM head pose encoding instead of depth map (ablation) | Face2Faceρ · Reenactment | FID: 45.1 vs 44.5 | No clear effect |
| use hierarchical coarse-to-fine motion network instead of single-scale network (ablation) | Face2Faceρ · Reenactment | AU-H: 0.32 -> 0.22 | Improved |
| use hierarchical coarse-to-fine motion network instead of single-scale network (ablation) | Face2Faceρ · Reconstruction | LPIPS: 0.131 -> 0.123 | Improved |
| use hierarchical motion network instead of single-scale hourglass (ablation) | motion module · inference speed | speedup: 3.5 times | Improved |
| support 16x downsampling with skip connections in rendering network (ablation) | rendering module · inference speed | speedup: 3 times faster | Improved |
| add VGGFace identity loss (ablation) | Face2Faceρ · identity preservation | CSIM | No clear effect |
| use SPADE instead of AdaIN for pose injection (ablation) | Face2Faceρ · Reenactment | quality | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- HeadGAN — HeadGAN: One-Shot Neural Head Synthesis and Editing (ICCV 2021) · better on 9 of 9 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Face2Face: Real-Time Face Capture and Reenactment of RGB Videos (CVPR 2016 · similarity 0.89)
- ReenactGAN: Learning to Reenact Faces via Boundary Transfer (ECCV 2018 · similarity 0.88)
- Learning Dense Correspondence for NeRF-Based Face Reenactment (AAAI 2024 · similarity 0.88)
- HyperReenact: One-Shot Reenactment via Jointly Learning to Refine and Retarget Faces (ICCV 2023 · similarity 0.88)
- ControlFace: Harnessing Facial Parametric Control for Face Rigging (CVPR 2025 · similarity 0.87)
- VOODOO 3D: Volumetric Portrait Disentanglement For One-Shot 3D Head Reenactment (CVPR 2024 · similarity 0.86)
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.
