Image Generation on LSUN bedroom
1.52FIDProjected GAN
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Projected GANResolution=256x2562021.11 | 1.52 | — | — | — | — | |
| Projected GANResolution=256x256, Training Point=First point surpassing SOTA2021.11 | 2.58 | — | — | — | — | |
| StyleGAN2Resolution=256x256, Training Point=Lowest FID in literature2021.11 | 2.65 | — | — | — | — | |
| FastGAN2021.11 | 3 | — | — | 2 | — | |
| StyleGAN2-ADA2021.11 | 5 | — | — | 10 | — | |
| F-PNDMsteps=20, variance schedule=linear2022.02 | 5.68 | — | — | — | — | |
| S-PNDMsteps=40, variance schedule=linear2022.02 | 5.74 | — | — | — | — | |
| F-PNDMsteps=25, variance schedule=linear2022.02 | 5.74 | — | — | — | — | |
| S-PNDMsteps=50, variance schedule=linear2022.02 | 5.81 | — | — | — | — | |
| DDIM*steps=100, variance schedule=linear2022.02 | 5.97 | — | — | — | — | |
| S-PNDMsteps=25, variance schedule=linear2022.02 | 6.02 | — | — | — | — | |
| DDIM*steps=125, variance schedule=linear2022.02 | 6.03 | — | — | — | — | |
| DDIM*steps=50, variance schedule=linear2022.02 | 6.05 | — | — | — | — | |
| GANsformersResolution=256x2562021.11 | 6.15 | — | — | — | — | |
| F-PNDMsteps=40, variance schedule=linear2022.02 | 6.17 | — | — | — | — | |
| DDIM*steps=200, variance schedule=linear2022.02 | 6.23 | — | — | — | — | |
| DDIM*steps=40, variance schedule=linear2022.02 | 6.27 | — | — | — | — | |
| S-PNDMsteps=100, variance schedule=linear2022.02 | 6.29 | — | — | — | — | |
| DDIM*steps=250, variance schedule=linear2022.02 | 6.32 | — | — | — | — | |
| S-PNDMsteps=125, variance schedule=linear2022.02 | 6.44 | — | — | — | — | |
| F-PNDMsteps=50, variance schedule=linear2022.02 | 6.44 | — | — | — | — | |
| S-PNDMsteps=20, variance schedule=linear2022.02 | 6.5 | — | — | — | — | |
| DDIMsteps=100, variance schedule=linear2022.02 | 6.62 | — | — | — | — | |
| S-PNDMsteps=200, variance schedule=linear2022.02 | 6.69 | — | — | — | — | |
| DDIMsteps=50, variance schedule=linear2022.02 | 6.75 | — | — | — | — | |
| S-PNDMsteps=250, variance schedule=linear2022.02 | 6.75 | — | — | — | — | |
| F-PNDMsteps=100, variance schedule=linear2022.02 | 6.91 | — | — | — | — | |
| F-PNDMsteps=250, variance schedule=linear2022.02 | 6.92 | — | — | — | — | |
| F-PNDMsteps=125, variance schedule=linear2022.02 | 6.96 | — | — | — | — | |
| F-PNDMsteps=10, variance schedule=linear2022.02 | 6.99 | — | — | — | — | |
| F-PNDMsteps=200, variance schedule=linear2022.02 | 7.03 | — | — | — | — | |
| DDIM*steps=25, variance schedule=linear2022.02 | 7.41 | — | — | — | — | |
| FastGANResolution=256x2562021.11 | 8.24 | — | — | — | — | |
| Progressive GANResolution=256x2562020.11 | 8.3 | — | — | — | — | |
| DDIM*steps=20, variance schedule=linear2022.02 | 8.47 | — | — | — | — | |
| DDIMsteps=20, variance schedule=linear2022.02 | 8.89 | — | — | — | — | |
| TTURarchitecture=WGAN-GP, learning rate (b, a)=3e-4, 1e-4, time (min)=19002017.06 | 9.5 | — | — | — | — | |
| S-PNDMsteps=10, variance schedule=linear2022.02 | 10.2 | — | — | — | — | |
| StyleGAN2-ADAResolution=256x2562021.11 | 11.53 | — | — | — | — | |
| F-PNDMsteps=5, variance schedule=linear2022.02 | 12.6 | — | — | — | — | |
| SSGANResolution=128x1282020.11 | 13.3 | — | — | — | — | |
| SAGANResolution=256x2562021.11 | 14.06 | — | — | — | — | |
| DC-VAEResolution=128x1282020.11 | 14.3 | — | — | — | — | |
| SNGANResolution=128x1282020.11 | 16 | — | — | — | — | |
| Projected GAN2021.11 | 16 | — | — | 21 | — | |
| DDIM*steps=10, variance schedule=linear2022.02 | 16.4 | — | — | — | — | |
| DDIMsteps=10, variance schedule=linear2022.02 | 17 | — | — | — | — | |
| StyleALAEResolution=256x2562020.11 | 17.13 | — | — | — | — | |
| S-PNDMsteps=5, variance schedule=linear2022.02 | 18.1 | — | — | — | — | |
| Original one time-scale update rulearchitecture=WGAN-GP, learning rate (b=a)=1e-4, time (min)=20102017.06 | 20.5 | — | — | — | — | |
| StackGAN-v22017.10 | 35.61 | 3.02 | 1.05 | — | — | |
| DDIM*steps=5, variance schedule=linear2022.02 | 51.3 | — | — | — | — | |
| TTURarchitecture=DCGAN, learning rate (b, a)=1e-5, 1e-4, updates=340k2017.06 | 57.5 | — | — | — | — | |
| Original one time-scale update rulearchitecture=DCGAN, learning rate (b=a)=5e-5, updates=70k2017.06 | 70.4 | — | — | — | — | |
| DATA2021.11 | 76 | — | — | 67 | — | |
| StackGAN-v12017.10 | 91.94 | 3.59 | 1.95 | — | — |