Research map / Image generation / Editing and style transfer
MFIM: Megapixel Facial Identity Manipulation
Sanghyeon Na
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 style maps to baseline MFIM (ablation) | MFIM baseline · CelebA-HQ | expression distance: 1.116 -> 0.398 | Improved |
| add style maps to baseline MFIM (ablation) | MFIM baseline · CelebA-HQ | pose-HN: 7.899 -> 3.795 | Improved |
| add style maps to baseline MFIM (ablation) | MFIM baseline · CelebA-HQ | identity distance: 70.160 -> 91.430 | Hurt |
| add shape loss L_shape to config B (ablation) | MFIM · CelebA-HQ | shape distance: 0.823 -> 0.635 | Improved |
| add shape loss L_shape to config B (ablation) | MFIM · CelebA-HQ | expression distance: 0.398 -> 0.864 | Hurt |
| add pose loss L_pose to config C (ablation) | MFIM · CelebA-HQ | pose distance: 0.085 -> 0.078 | Improved |
| add expression loss L_exp to config D (ablation) | MFIM · CelebA-HQ | expression distance: 0.860 -> 0.400 | Improved |
| add expression loss L_exp to config D (ablation) | MFIM · CelebA-HQ | shape distance: 0.634 -> 0.782 | Hurt |
| add landmark loss L_lm to config B (ablation) | MFIM · CelebA-HQ (30k pairs) | shape distance: 0.842 -> 0.892 | No clear effect |
| use parameter loss L_param instead of landmark loss L_lm on config B (ablation) | MFIM · CelebA-HQ (30k pairs) | shape distance: 0.892 -> 0.842 | Improved |
| use parameter loss L_param instead of landmark loss L_lm on config B (ablation) | MFIM · CelebA-HQ (30k pairs) | pose distance: 0.046 -> 0.060 | Hurt |
| add partial-keypoint landmark loss L_lm to config E (ablation) | MFIM · CelebA-HQ | pose distance: 0.057 -> 0.041 | Improved |
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.
- SmoothSwap — Smooth-Swap: A Simple Enhancement for Face-Swapping With Smoothness (CVPR 2022) · better on 2 of 2 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
- One Shot Face Swapping on Megapixels (CVPR 2021 · similarity 0.90)
- StyleFace: Towards Identity-Disentangled Face Generation on Megapixels (ECCV 2022 · similarity 0.89)
- DynamicFace: High-Quality and Consistent Face Swapping for Image and Video using Composable 3D Facial Priors (ICCV 2025 · similarity 0.88)
- High-Fidelity Diffusion Face Swapping with ID-Constrained Facial Conditioning (CVPR 2026 · similarity 0.88)
- BlendFace: Re-designing Identity Encoders for Face-Swapping (ICCV 2023 · similarity 0.88)
- When StyleGAN Meets Stable Diffusion: a W+ Adapter for Personalized Image Generation (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.
