Research map / Image generation / Editing and style transfer
Attribute-Preserving Face Dataset Anonymization via Latent Code Optimization
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 |
|---|---|---|---|
| Use latent-code optimization with the full attribute-preserving method rather than CIAGAN | Face anonymization · CelebA-HQ | Combined attribute classification accuracy: 0.7852 -> 0.8181 | Helped |
| Use latent-code optimization with the full attribute-preserving method rather than DeepPrivacy | Face anonymization · CelebA-HQ | Combined attribute classification accuracy: 0.8135 -> 0.8181 | Helped |
| Use the full method rather than CIAGAN | Face anonymization · CelebA-HQ | FID: 37.94 -> 29.93 | Helped |
| Use the full method rather than DeepPrivacy | Face anonymization · CelebA-HQ | MTCNN detection rate: 99.85 -> 100 | Helped |
| Use the full method rather than CIAGAN | Face anonymization · LFW | MTCNN detection rate: 99.89 -> 100 | Helped |
| Use the full method rather than CIAGAN | Face anonymization · LFW (CelebA-HQ target distribution) | FID: 85.23 -> 68.88 | Helped |
| Add the attribute-preservation loss (ablation) | Ours (ID) face anonymization · CelebA-HQ | MTCNN detection rate: 97.99 -> 100 | Helped |
| Use margin m=.9 instead of m=.0 (ablation) | Ours face anonymization · CelebA-HQ | Attribute classification accuracy: 0.8181 -> 0.83 | Helped |
| Use margin m=.9 instead of m=.0 (ablation) | Ours face anonymization · CelebA-HQ | CASIA face re-ID rate: 2.80 -> 3.41 | Hurt |
| Use margin m=.9 instead of m=.0 (ablation) | Ours face anonymization · CelebA-HQ | FID: 29.93 -> 27.58 | Helped |
| Use patch-level FaRL features rather than the CLS token for attribute matching (ablation) | Ours latent-code optimization · Face attribute preservation | Attribute preservation | Helped |
Closest papers by meaning
- Disguise without Disruption: Utility-Preserving Face De-identification (AAAI 2024 · similarity 0.88)
- Balancing Privacy and Performance: A Many-in-One Approach for Image Anonymization (AAAI 2025 · similarity 0.88)
- Enhancing Facial Privacy Protection via Weakening Diffusion Purification (CVPR 2025 · similarity 0.87)
- Vec2Face: Scaling Face Dataset Generation with Loosely Constrained Vectors (ICLR 2025 · similarity 0.87)
- FaceA-Net: Facial Attribute-Driven ID Preserving Image Generation Network (AAAI 2025 · similarity 0.87)
- How Private is Private? A Comparative Study for Face De-Identification (NeurIPS 2026 · similarity 0.87)
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.