Research map / Image generation / Editing and style transfer
Realistic Face Reconstruction from Facial Embeddings via Diffusion Models
Dong Han, Yong Li, Joachim Denzler
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 |
|---|---|---|---|
| replace MLP mapping with KAN mapping in FEM (ablation) | FEM · CelebA-HQ, IR152 | ASR average (%): 71.4 -> 77.2 | Improved |
| replace MLP mapping with KAN mapping in FEM (ablation) | FEM · CelebA-HQ, IRSE50 | ASR average (%): 81.5 -> 83.7 | Improved |
| replace MLP mapping with KAN mapping in FEM (ablation) | FEM · CelebA-HQ, MinusFace | ASR average (%): 67.1 -> 67.6 | No clear effect |
| replace MLP mapping with KAN mapping in FEM (ablation) | FEM · LADN-M makeup | ASR average (%): 81.1 -> 85.1 | Improved |
| add the FEM embedding mapping before IPA-FaceID (ablation) | IPA-FaceID · CelebA-HQ, IRSE50 | ASR average (%): 7.5 -> 83.7 | Improved |
| apply makeup to the face images used for embeddings | FaceTI · LADN | ASR average (%): 74.5 -> 56.4 | Hurt |
| apply makeup to the face images used for embeddings | MAP2V · LADN | ASR average (%): 84.3 -> 75.5 | Hurt |
| apply makeup to the face images used for embeddings | FEM-KAN · LADN | ASR average (%): 91.5 -> 85.1 | Hurt |
| use FEM instead of MAP2V on HFCF PPFR embeddings | FEM-KAN · CelebA-HQ, HFCF | ASR average (%): 28.0 -> 83.1 | Improved |
| reduce leaked embedding fraction (partial leakage) for reconstruction | FEM-KAN · IRSE50 partial embeddings, ArcFace | ASR (%): 32.5% ASR below 30% leakage | Hurt |
| cloak images with Fawkes before extracting embeddings | FEM-KAN · CelebA-HQ IRSE50, ElasticFace | ASR (%): 93.8 -> 47.1 | Hurt |
| use FEM instead of FaceTI for training | mapping training cost · 90% FFHQ | training time per epoch: 51 hrs -> 3 hrs | Improved |
Closest papers by meaning
- Enhancing Facial Privacy Protection via Weakening Diffusion Purification (CVPR 2025 · similarity 0.89)
- $\text{ID}^3$: Identity-Preserving-yet-Diversified Diffusion Models for Synthetic Face Recognition (NeurIPS 2024 · similarity 0.87)
- Validating Privacy-Preserving Face Recognition under a Minimum Assumption (CVPR 2024 · similarity 0.87)
- Transferable Facial Privacy Protection against Blind Face Restoration via Domain-Consistent Adversarial Obfuscation (ICML 2024 · similarity 0.87)
- Privacy-Preserving Face Recognition Using Trainable Feature Subtraction (CVPR 2024 · similarity 0.86)
- FSFM: A Generalizable Face Security Foundation Model via Self-Supervised Facial Representation Learning (CVPR 2025 · 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.
