Model Inversion Attack on CelebA (private) FFHQ (public) on IR152 (test)
92Top-5 Attack AccuracyKEDMI + LOMMA
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| KEDMI + LOMMATarget Model=IR152, Evaluation Model=face.evoLve, Public Dataset=FFHQ, Private Dataset=CelebA2023.04 | 92 | 6.6 | 45.67 | |
| GMI + LOMMATarget Model=IR152, Evaluation Model=face.evoLve, Public Dataset=FFHQ, Private Dataset=CelebA2023.04 | 90.33 | 54 | 37.58 | |
| KEDMI + LOMTarget Model=IR152, Evaluation Model=face.evoLve, Public Dataset=FFHQ, Private Dataset=CelebA2023.04 | 88.67 | 3.33 | 50.84 | |
| KEDMI + MATarget Model=IR152, Evaluation Model=face.evoLve, Public Dataset=FFHQ, Private Dataset=CelebA2023.04 | 87.67 | 2.33 | 39.88 | |
| KEDMITarget Model=IR152, Evaluation Model=face.evoLve, Public Dataset=FFHQ, Private Dataset=CelebA2023.04 | 85.33 | — | 41.71 | |
| GMI + MATarget Model=IR152, Evaluation Model=face.evoLve, Public Dataset=FFHQ, Private Dataset=CelebA2023.04 | 84 | 47.67 | 35.41 | |
| GMI + LOMTarget Model=IR152, Evaluation Model=face.evoLve, Public Dataset=FFHQ, Private Dataset=CelebA2023.04 | 80.33 | 44 | 40.18 | |
| GMITarget Model=IR152, Evaluation Model=face.evoLve, Public Dataset=FFHQ, Private Dataset=CelebA2023.04 | 36.33 | — | 47.72 |