Image Generation on FFHQ 256x256 (train)
2.96FIDEqGAN-SA
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| EqGAN-SAResolution=256x256, Number of Training Images=25M, Alignment Loss (Lalign)=true2021.12 | 2.96 | — | — | — | — | — | — | — | 0.73 | |
| Projected GANResolution=256x2562021.11 | 3.08 | 65.4 | 46.4 | 0.0004 | 0.85 | 7.55 | 0.64 | 6,410 | — | |
| EqGAN-SA (SEL)Resolution=256x256, Number of Training Images=25M2021.12 | 3.39 | — | — | — | — | — | — | — | 1.38 | |
| StyleGAN2Resolution=256x256, Number of Training Images=25M2021.12 | 3.66 | — | — | — | — | — | — | — | 1.62 | |
| SWAGAN-BiImages Viewed=8.5M2021.02 | 5.22 | — | — | — | — | — | — | — | — | |
| StyleGAN2Images Viewed=8.5M2021.02 | 5.82 | — | — | — | — | — | — | — | — | |
| StyleGAN22021.12 | 6.97 | — | — | 0.17 | — | — | — | — | — | |
| StyleGAN2-ADAResolution=256x2562021.11 | 7.32 | 66.9 | 44.5 | 0.0015 | 1.21 | 10.3 | 1.2 | 7,420 | — | |
| Wavelet at final layerImages Viewed=8.5M2021.02 | 7.96 | — | — | — | — | — | — | — | — | |
| SWAGAN-NUImages Viewed=8.5M2021.02 | 9.2 | — | — | — | — | — | — | — | — | |
| Dia-StyleGAN2architecture=StyleGAN2, sampling_method=Dia2021.02 | 11.89 | 69 | 30 | — | — | — | — | — | — | |
| SWAGAN-Bi w/ NWDImages Viewed=8.5M2021.02 | 12.27 | — | — | — | — | — | — | — | — | |
| FastGANResolution=256x2562021.11 | 12.69 | 71.6 | 18.4 | 0.0053 | 2.55 | 19.23 | 2.47 | 10,190 | — | |
| StyleGAN2architecture=StyleGAN22021.02 | 14.07 | 72 | 27 | — | — | — | — | — | — | |
| GRAMBatch size=32, Training iterations=150K2021.12 | 14.5 | — | — | 0.65 | — | — | — | — | — | |
| GOLDarchitecture=StyleGAN2, sampling_method=GOLD2021.02 | 15.53 | 69 | 29 | — | — | — | — | — | — | |
| GRAM2021.12 | 17.9 | — | — | 0.84 | — | — | — | — | — | |
| GIRAFFEEvaluated using pre-trained models provided by the authors=true2021.12 | 32.6 | — | — | 2.24 | — | — | — | — | — | |
| pi-GANEvaluated using pre-trained models provided by the authors=true2021.12 | 55.2 | — | — | 4.13 | — | — | — | — | — | |
| GRAFEvaluated using pre-trained models provided by the authors=true2021.12 | 73 | — | — | 5.89 | — | — | — | — | — |