Research map / Label-efficient and robust learning / Recognition and long-tail
FlowFace: Semantic Flow-Guided Shape-Aware Face Swapping
Read the paperCheck 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 |
|---|---|---|---|
| Add semantic-flow face reshaping before SimSwap (ablation) | SimSwap · FF++ | ID Acc (Avg): 93.63% to 94.31% | Helped |
| Add semantic-flow face reshaping to the face-swapping stage (ablation) | FlowFace · FF++ | Shape: 1.43 -> 1.17 | Helped |
| Remove segmentation input from the semantic-flow generator (ablation) | Face reshaping network · Face reshaping | flow accuracy | Hurt |
| Add segmentation input to the semantic-flow generator (ablation) | Face reshaping network · Face reshaping | flow accuracy | Helped |
| Remove semantic inputs from the reshaping discriminator (ablation) | Face reshaping network · Face reshaping | visual naturalness | Hurt |
| Remove the reshaping discriminator (ablation) | Face reshaping network · Face reshaping | visual quality | Hurt |
| Replace cross-attention fusion with direct addition (ablation) | FlowFace face-swapping network · FF++ | Pose: 1.99 -> 4.90 | Hurt |
| Replace cross-attention fusion with AdaIN (ablation) | FlowFace face-swapping network · FF++ | ID Acc (Avg): 98.71 -> 97.23 | Hurt |
| Use MAE latent features instead of an identity embedding, with cross-attention fusion (ablation) | FlowFace face-swapping network · FF++ | ID Acc (Avg): 97.00 -> 98.71 | Helped |
| Use a convolutional decoder instead of a ViT decoder (ablation) | FlowFace face-swapping network · FF++ | ID Acc (Avg): 98.09 -> 98.71 | Helped |
| Use FlowFace rather than HifiFace | Face swapping · FF++ | Pose: 2.29 -> 2.66 | Hurt |
Closest papers by meaning
- Fine-Grained Face Swapping via Regional GAN Inversion (CVPR 2023 · similarity 0.90)
- Instance-Level Facial Attributes Transfer with Geometry-Aware Flow (AAAI 2019 · similarity 0.90)
- Zero-Shot Head Swapping in Real-World Scenarios (CVPR 2025 · similarity 0.86)
- ControlFace: Harnessing Facial Parametric Control for Face Rigging (CVPR 2025 · similarity 0.86)
- When StyleGAN Meets Stable Diffusion: a W+ Adapter for Personalized Image Generation (CVPR 2024 · similarity 0.85)
- Enhancing Identity-Deformation Disentanglement in StyleGAN for One-Shot Face Video Re-Enactment (AAAI 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.