Image Generation on CAT 64x64 (N=9304)
15.61Mean FIDRaLSGAN
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| RaLSGAN2018.07 | 15.61 | 11.97 | 19.29 | 2.55 | |
| HingeGANGradient Penalty Function=max(0, ||∇xf(x)||_∞ - 1), Generator Iterations=100k, Architecture=standard CNN2019.10 | 17.5 | — | — | — | |
| RSGAN-GP2018.07 | 18.2 | 16.41 | 22.34 | 1.82 | |
| WGANGradient Penalty Function=max(0, ||∇xf(x)||_∞ - 1), Generator Iterations=100k, Architecture=standard CNN2019.10 | 18.3 | — | — | — | |
| RaSGAN-GP2018.07 | 19.58 | 17.32 | 22 | 1.81 | |
| RaHingeGAN2018.07 | 20.29 | 14.62 | 27.31 | 3.96 | |
| RaSGAN2018.07 | 20.53 | 15.38 | 33.11 | 5.68 | |
| HingeGANGradient Penalty Function=(||∇xf(x)||_2 - 1)^2, Generator Iterations=100k, Architecture=standard CNN2019.10 | 26.7 | — | — | — | |
| HingeGANGradient Penalty Function=max(0, ||∇xf(x)||_2 - 1), Generator Iterations=100k, Architecture=standard CNN2019.10 | 29.6 | — | — | — | |
| RSGAN2018.07 | 32.16 | 19.03 | 42.05 | 7.01 | |
| HingeGAN2018.07 | 32.23 | 17.6 | 50.94 | 14.44 | |
| WGANGradient Penalty Function=max(0, ||∇xf(x)||_2 - 1), Generator Iterations=100k, Architecture=standard CNN2019.10 | 43.7 | — | — | — | |
| WGANGradient Penalty Function=(||∇xf(x)||_2 - 1)^2, Generator Iterations=100k, Architecture=standard CNN2019.10 | 48.2 | — | — | — | |
| SGAN2018.07 | 52.54 | 16.56 | 310.56 | 96.81 | |
| LSGAN2018.07 | 73.62 | 20.27 | 224.97 | 61.02 |