Unsupervised Image Generation on STL-10 (train)
26.08Inception ScoreReal data
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Real data2018.02 | 26.08 | 7.9 | |
| SN-GANsBackbone Architecture=ResNet, Optimization variant=Eq.(17)2018.02 | 9.1 | 40.1 | |
| SN-GANsBackbone Architecture=Standard CNN, Update schedule=2x updates, Optimization variant=Eq.(17)2018.02 | 8.79 | 43.2 | |
| OrthonormalBackbone Architecture=ResNet, Optimization variant=Eq.(17)2018.02 | 8.72 | 42.4 | |
| SN-GANsBackbone Architecture=Standard CNN, Update schedule=2x updates2018.02 | 8.69 | 47.5 | |
| OrthonormalBackbone Architecture=Standard CNN, Update schedule=2x updates2018.02 | 8.67 | 44.2 | |
| OrthonormalBackbone Architecture=Standard CNN2018.02 | 8.56 | 46.7 | |
| Warde-Farley et al.2018.02 | 8.51 | — | |
| WGAN-GPBackbone Architecture=Standard CNN2018.02 | 8.42 | 55.1 | |
| SN-GANsBackbone Architecture=Standard CNN2018.02 | 8.28 | 53.1 | |
| DCGAN2018.02 | 7.84 | — | |
| Layer Norm.Backbone Architecture=Standard CNN2018.02 | 7.61 | 75.6 | |
| Weight clippingBackbone Architecture=Standard CNN2018.02 | 7.57 | 64.2 | |
| Weight Norm.Backbone Architecture=Standard CNN2018.02 | 7.16 | 73.4 |