Image Reconstruction on ImageNet1K
15.515PSNRGradViT
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| GradViTNetwork=ViT-B/16, Type=ViT, Need Original Labels=No, GAN-Based=No2022.03 | 15.515 | 0.032 | 0.295 | |
| GradViTNetwork=DeiT-B/16, Type=ViT, Need Original Labels=No, GAN-Based=No2022.03 | 13.252 | 0.058 | 0.413 | |
| GradInversionNetwork=ResNet-50, Type=CNN, Need Original Labels=No, GAN-Based=No2022.03 | 12.929 | 0.175 | 0.484 | |
| Inverting GradientsNetwork=ResNet-50, Type=CNN, Need Original Labels=Yes, GAN-Based=No2022.03 | 11.703 | 0.355 | 0.749 | |
| GradViTNetwork=ResNet-50, Type=CNN, Need Original Labels=No, GAN-Based=No2022.03 | 11.635 | 0.076 | 0.454 | |
| GradInversionNetwork=ViT-B/16, Type=ViT, Need Original Labels=No, GAN-Based=No2022.03 | 10.824 | 0.116 | 0.708 | |
| Deep Gradient LeakageNetwork=ResNet-50, Type=CNN, Need Original Labels=No, GAN-Based=No2022.03 | 10.252 | 1.319 | 0.602 | |
| Latent projectionNetwork=BigGAN, Type=CNN, Need Original Labels=Yes, GAN-Based=Yes2022.03 | 10.149 | 0.275 | 0.722 | |
| DeepInversionNetwork=ResNet-50, Type=CNN, Need Original Labels=Yes, GAN-Based=No2022.03 | 10.131 | 0.238 | 0.728 | |
| Random NoiseNetwork=null, Type=null, Need Original Labels=No, GAN-Based=No2022.03 | 1.351 | 0.706 | 9.964 |