Model Inversion on CelebA
74.9Attack AccuracyVMI + LOMMA
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| VMI + LOMMATarget Model=ResNet-34, Evaluation Model=IR-SE50, Public Dataset (Dpub)=CelebA, Private Dataset (Dpriv)=CelebA2023.04 | 74.9 | 14.94 | 1.109 | |
| VMI + LOMTarget Model=ResNet-34, Evaluation Model=IR-SE50, Public Dataset (Dpub)=CelebA, Private Dataset (Dpriv)=CelebA2023.04 | 68.34 | 8.38 | 1.131 | |
| VMI + MATarget Model=ResNet-34, Evaluation Model=IR-SE50, Public Dataset (Dpub)=CelebA, Private Dataset (Dpriv)=CelebA2023.04 | 64.16 | 4.2 | 1.14 | |
| VMITarget Model=ResNet-34, Evaluation Model=IR-SE50, Public Dataset (Dpub)=CelebA, Private Dataset (Dpriv)=CelebA2023.04 | 59.96 | — | 1.144 |