Research map / Image generation / Editing and style transfer
Machine Pareidolia: Protecting Facial Image with Emotional Editing
Binh M. Le, Simon S. Woo
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 gradient projection between identity and emotion losses (ablation) | MAP · IR152, CelebA-HQ identity 1 | PSR (FAR@1e-2): 71.2 -> 80.3 | Improved |
| add gradient projection (ablation) | MAP · IR152, CelebA-HQ identity 1 | FID: 24.5 -> 21.5 | Improved |
| add EMA gradient estimate (ablation) | MAP · IR152, CelebA-HQ identity 1 | PSR (FAR@1e-2): 76.5 -> 80.3 | Improved |
| add EMA gradient estimate (ablation) | MAP · IR152, CelebA-HQ identity 1 | FID: 22.5 -> 21.5 | Improved |
| add Laplacian smoothness loss (ablation) | MAP · IR152, CelebA-HQ identity 1 | PSR (FAR@1e-2): 78.8 -> 80.3 | Improved |
| add Laplacian smoothness loss (ablation) | MAP · IR152, CelebA-HQ identity 1 | FID: 23.6 -> 21.5 | Improved |
| add emotion editing (ablation) | MAP · IR152, CelebA-HQ identity 1 | PSR (FAR@1e-2): 85.0 -> 80.3 | Hurt |
| add emotion editing (ablation) | MAP · IR152, CelebA-HQ identity 1 | FID: 24.1 -> 21.5 | Improved |
| set score matching weight gamma_d to 0 (remove score matching) (ablation) | MAP · CelebA-HQ PSR vs FID trade-off | PSR/FID trade-off | Hurt |
| replace DiffAM makeup baseline with MAP emotion editing | face protection · CelebA-HQ and LADN average | PSR (FAR@1e-2): 77.88 -> 88.48 | Improved |
| use MAP instead of Adv-CPG for identification | face identification · CelebA-HQ IRSE50 | Rank-1 PSR: 24.4 -> 57.9 | Improved |
| use MAP instead of DiffAM | face protection · Image quality average | FID: 26.10 -> 24.21 | 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.
- CLIP2Protect — CLIP2Protect: Protecting Facial Privacy Using Text-Guided Makeup via Adversarial Latent Search (CVPR 2023) · better on 1 of 1 shared result
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
- Edit Away and My Face Will not Stay: Personal Biometric Defense against Malicious Generative Editing (CVPR 2025 · similarity 0.87)
- How Private is Private? A Comparative Study for Face De-Identification (NeurIPS 2026 · similarity 0.87)
- MYOPIA: Protecting Face Privacy from Malicious Personalized Text-to-Image Synthesis via Unlearnable Examples (AAAI 2025 · similarity 0.87)
- Disguise without Disruption: Utility-Preserving Face De-identification (AAAI 2024 · similarity 0.86)
- Validating Privacy-Preserving Face Recognition under a Minimum Assumption (CVPR 2024 · similarity 0.86)
- Privacy-Preserving Optics for Enhancing Protection in Face De-Identification (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.
