Facescrub
Benchmarks
| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| PPA Model Inversion Attack | FaceScrub 224x224 Dpriv = Facescrub, Dpub = FFHQ (test) | Accuracy98.02 | 27 | |
| Face Identification | MF1-Facescrub 1.0 (test) | Rank-1 Identification Rate98.71 | 26 | |
| Model Inversion Defense | Facescrub 224x224 (test) | Accuracy98.46 | 21 | |
| Image Reconstruction | FaceScrub (val) | SSIM (Dec-based MIA)89.8 | 16 | |
| Image Classification | FaceScrub (val) | Accuracy85.12 | 16 | |
| Image Classification | Facescrub (inference) | Accuracy86.69 | 15 | |
| Image Classification | Facescrub (train) | Dec.-based MSE0.0012 | 15 | |
| Model Inversion Attack | Facescrub | Attack Accuracy58.6 | 10 | |
| Face Recognition Attack | FaceScrub (test) | IR152 Score73.14 | 7 | |
| Model Inversion Attack Defense | Facescrub | Accuracy97.9 | 6 | |
| Model Inversion Defense | FaceScrub (test) | Accuracy96.57 | 6 | |
| User study on image reconstruction similarity | FaceScrub 530 classes | Number of Similar Selections537 | 4 | |
| Model Inversion Attack | Facescrub (private identities) | Attack Accuracy0.48 | 4 | |
| Membership Inference Defense | FaceScrub | Accuracy96.57 | 2 | |
| Image Classification | FACESCRUB (test) | Accuracy (Test)82.7 | 2 | |
| Face Identity Retrieval | Facescrub 100,001 face images (test) | Median Rank16 | 2 |
Showing 16 of 16 rows