Image Generation on LSUN Cat 10k samples (train)
1.78SwAV FIDVision-aided Adversarial training
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Vision-aided Adversarial trainingAugmentation Strategy=ADA, Fine-tuning=true2021.12 | 1.78 | — | |
| Vision-aided Adversarial trainingAugmentation Strategy=DiffAugment, Fine-tuning=true2021.12 | 1.96 | — | |
| ADAAugmentation Strategy=ADA2021.12 | 4.5 | — | |
| DiffAugmentAugmentation Strategy=DiffAugment2021.12 | 4.97 | — | |
| StyleGAN2Augmentation Strategy=none2021.12 | 6.92 | — | |
| ADA2020.06 | — | 13.13 | |
| StyleGAN22020.06 | — | 17.93 | |
| StyleGAN2 + DiffAugmentAugmentation=DiffAugment (Color + Translation + Cutout)2020.06 | — | 12.07 |