Research map / Image generation / Editing and style transfer
High-Resolution Face Swapping via Latent Semantics Disentanglement
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 |
|---|---|---|---|
| Disentangle structure and appearance in StyleGAN latent codes, using landmark-guided structure transfer and target appearance codes | Face swapping with StyleGAN2 · CelebA-HQ | ID similarity: 0.5214 -> 0.5688 | Helped |
| Disentangle structure and appearance in StyleGAN latent codes, using landmark-guided structure transfer and target appearance codes | Face swapping with StyleGAN2 · CelebA-HQ | FID: 11.645 -> 9.987 | Helped |
| Disentangle structure and appearance in StyleGAN latent codes, using landmark-guided structure transfer and target appearance codes | Face swapping with StyleGAN2 · FaceForensics++ | ID retrieval rate (%): 90.83 -> 90.05 | Mixed |
| Disentangle structure and appearance in StyleGAN latent codes, using landmark-guided structure transfer | Face swapping with StyleGAN2 · FaceForensics++ | Pose error: 2.64 -> 2.46 | Helped |
| Add target appearance-code swapping and the style-transfer loss (ablation) | Var.1 face-swapping variant · CelebA-HQ | FID: 10.986 -> 9.987 | Helped |
| Add target appearance-code swapping and the style-transfer loss (ablation) | Var.1 face-swapping variant · CelebA-HQ | ID similarity: 0.5645 -> 0.5688 | No clear effect |
| Add the multi-resolution background-transfer module (ablation) | Var.2 face-swapping variant · CelebA-HQ | FID: 10.345 -> 9.987 | Helped |
| Add the multi-resolution background-transfer module (ablation) | Var.2 face-swapping variant · CelebA-HQ | Pose error: 3.023 -> 2.997 | No clear effect |
| Use a landmark encoder rather than an image encoder for the structure-transfer direction (ablation) | Var.3 face-swapping variant · CelebA-HQ | Pose error: 3.242 -> 2.997 | Helped |
| Use a landmark encoder rather than an image encoder for the structure-transfer direction (ablation) | Var.3 face-swapping variant · CelebA-HQ | Expression error: 3.11 -> 2.74 | Helped |
| Add the code trajectory constraint | Video face swapping · High-resolution target videos | Temporal coherence | Helped |
| Add the flow trajectory constraint | Video face swapping · High-resolution target videos | Temporal coherence | Helped |
Closest papers by meaning
- Interpreting the Latent Space of GANs for Semantic Face Editing (CVPR 2020 · similarity 0.91)
- Fine-Grained Face Swapping via Regional GAN Inversion (CVPR 2023 · similarity 0.91)
- 3D-Aware Face Swapping (CVPR 2023 · similarity 0.90)
- VividFace: A Robost and High-Fidelity Video Face Swapping Framework (NeurIPS 2025 · similarity 0.89)
- High-Fidelity Diffusion Face Swapping with ID-Constrained Facial Conditioning (CVPR 2026 · similarity 0.88)
- DiffSwap: High-Fidelity and Controllable Face Swapping via 3D-Aware Masked Diffusion (CVPR 2023 · similarity 0.88)
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.