Research map / Image generation / Editing and style transfer
Visible Yet Unrecognizable: Frequency-Selective Facial Privacy via Attention
Atul Kumar, Akshay Agarwal
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 cross-attention identity perturbation on LL subband (ablation) | MIRAGE (AdaFace) · CelebA-HQ proactive | recognition accuracy: 96.30 -> 70.24 | Improved |
| Add self-attention on high-frequency subbands (ablation) | MIRAGE (AdaFace) · CelebA-HQ proactive | recognition accuracy: 92.20 -> 70.24 | Improved |
| Add FAC loss (ablation) | MIRAGE (AdaFace) · CelebA-HQ proactive | recognition accuracy: 92.98 -> 70.24 | Improved |
| Use learned attention instead of naive DWT sub-band swapping (ablation) | MIRAGE (AdaFace) · CelebA-HQ proactive | recognition accuracy: 95.47 -> 70.24 | Improved |
| Apply JPEG compression to protected images | MIRAGE (AdaFace) · CelebA-HQ | recognition accuracy: 70.24 -> 64.02 | Improved |
| Fine-tune FR model on MIRAGE-protected images | AdaFace · CelebA-HQ | recognition accuracy: +1.84% | No clear effect |
| Use MIRAGE instead of FALCO for privacy strength | FALCO · CelebA-HQ AdaFace proactive | recognition accuracy: 56.77 vs 70.24 | Hurt |
| Use MIRAGE instead of FPP/DiffAM | FPP / DiffAM · CelebA-HQ AdaFace proactive | recognition accuracy: 96.19 / 96.39 -> 70.24 | Improved |
Closest papers by meaning
- Disguise without Disruption: Utility-Preserving Face De-identification (AAAI 2024 · similarity 0.89)
- How Private is Private? A Comparative Study for Face De-Identification (NeurIPS 2026 · similarity 0.88)
- Privacy-Preserving Face Recognition Using Trainable Feature Subtraction (CVPR 2024 · similarity 0.88)
- Privacy-Preserving Optics for Enhancing Protection in Face De-Identification (CVPR 2024 · similarity 0.87)
- Validating Privacy-Preserving Face Recognition under a Minimum Assumption (CVPR 2024 · similarity 0.87)
- PerceptAnon: Exploring the Human Perception of Image Anonymization Beyond Pseudonymization for GDPR (ICML 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.
