Image Generation on FFHQ
2.02FIDStyleGAN-XL
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| StyleGAN-XLInput=Gaussian2023.05 | 2.02 | — | |
| Spider StyleGAN2-ADAInput=TinyImageNet2023.05 | 2.45 | — | |
| Pixel DiffusionRepresentation Space=Pixel, Architecture=U-Net, Framework=EDM2026.04 | 2.56 | — | |
| Polarity-StyleGAN2Input=Gaussian2023.05 | 2.57 | — | |
| MaGNET-StyleGAN2Input=Gaussian2023.05 | 2.66 | — | |
| StyleGAN2-ADAInput=Gaussian2023.05 | 2.7 | — | |
| StyleGAN3-TInput=Gaussian2023.05 | 2.79 | — | |
| Spider StyleGAN3-TInput=TinyImageNet2023.05 | 2.86 | — | |
| Spider StyleGAN2-ADAInput=AFHQ-Dogs2023.05 | 3.07 | — | |
| INDISNFE=82026.03 | 3.31 | — | |
| Projected GANResolution=256x2562021.11 | 3.39 | — | |
| Projected GANResolution=256x256, Training Point=First point surpassing SOTA2021.11 | 3.56 | — | |
| StyleGAN2Resolution=256x256, Training Point=Lowest FID in literature2021.11 | 3.62 | — | |
| StyleGAN2Type=GAN2022.04 | 3.73 | — | |
| StyleGANv2Resolution=256x2562020.11 | 3.83 | — | |
| INDISNFE=72026.03 | 3.9 | — | |
| LD3NFE=82026.03 | 3.97 | — | |
| CIPSResolution=256x256, Note=Trained for more time and with larger batch size than in Tab. 32020.11 | 4.38 | — | |
| StyleGAN2Resolution=1024x1024, Image number=70k2021.01 | 4.4 | — | |
| OriginalModel=StyleGAN2 [66], MACs=45.1B2023.03 | 4.5 | — | |
| LD3NFE=72026.03 | 4.69 | — | |
| INDISNFE=62026.03 | 4.72 | — | |
| GITSNFE=82026.03 | 4.91 | — | |
| StyleMatteGANvision-aided=true2023.05 | 5.13 | — | |
| GITSNFE=72026.03 | 5.36 | — | |
| LD3NFE=62026.03 | 5.97 | — | |
| GITSNFE=62026.03 | 6.39 | — | |
| Labels4Free2023.05 | 6.51 | — | |
| AMEDNFE=72026.03 | 6.64 | — | |
| best heu.NFE=82026.03 | 6.86 | — | |
| INDISNFE=52026.03 | 6.91 | — | |
| P2 weightingType=Diff, Sampling Steps=10002022.04 | 6.92 | — | |
| P2 weightingType=Diff, Sampling Steps=5002022.04 | 6.97 | — | |
| StyleGAN2-ADAResolution=256x2562021.11 | 7.32 | — | |
| GANsformersResolution=256x2562021.11 | 7.42 | — | |
| CAGC + VEMModel=StyleGAN2 [66], MACs=4.1B, Compression rate=90.91%2023.03 | 7.48 | — | |
| StyleGAN2Resolution=1024x1024, Image number=10k2021.01 | 7.86 | — | |
| AMEDNFE=82026.03 | 7.86 | — | |
| CAGCModel=StyleGAN2 [66], MACs=4.1B, Compression rate=90.91%2023.03 | 7.9 | — | |
| StyleMatteGANvision-aided=false2023.05 | 8.21 | — | |
| Diffusion BaselineType=Diff, Sampling Steps=5002022.04 | 8.41 | — | |
| LD3NFE=52026.03 | 8.56 | — | |
| FurryGAN2023.05 | 8.72 | — | |
| best heu.NFE=72026.03 | 8.76 | — | |
| GITSNFE=52026.03 | 8.85 | — | |
| INDISNFE=42026.03 | 8.92 | — | |
| VQGANType=GAN+AR2022.04 | 9.6 | — | |
| AMEDNFE=62026.03 | 9.97 | — | |
| DMNNFE=72026.03 | 10.17 | — | |
| StyleGAN2Resolution=1024x1024, Image number=5k2021.01 | 10.45 | — | |
| Factorized Diffusion ArchitecturesResolution=1282023.09 | 10.79 | — | |
| UNet GANType=GAN2022.04 | 10.9 | — | |
| DMNNFE=82026.03 | 11 | — | |
| best heu.NFE=62026.03 | 11.44 | — | |
| Factorized Diffusion ArchitecturesResolution=642023.09 | 12.02 | — | |
| DMNNFE=62026.03 | 12.12 | — | |
| FastGANResolution=1024x1024, Image number=70k2021.01 | 12.38 | — | |
| BigGANType=GAN2022.04 | 12.4 | — | |
| GAN-SlimmingModel=StyleGAN2 [66], MACs=5.0B, Compression rate=88.91%2023.03 | 12.4 | — | |
| AMEDNFE=52026.03 | 12.49 | — | |
| FastGANResolution=256x2562021.11 | 12.69 | — | |
| D2CType=Diff2022.04 | 13.04 | — | |
| DDPMResolution=1282023.09 | 13.35 | — | |
| GITSNFE=42026.03 | 13.59 | — | |
| DDPMResolution=642023.09 | 13.72 | — | |
| LD3NFE=42026.03 | 14.15 | — | |
| SAGANResolution=256x2562021.11 | 16.21 | — | |
| FastGANResolution=1024x1024, Image number=10k2021.01 | 16.45 | — | |
| best heu.NFE=52026.03 | 17.52 | — | |
| BaselineResolution=1024x1024, Image number=70k2021.01 | 17.62 | — | |
| INDISNFE=32026.03 | 17.72 | — | |
| FastGANResolution=1024x1024, Image number=5k2021.01 | 17.93 | — | |
| StyleGAN2Resolution=1024x1024, Image number=2k2021.01 | 18.38 | — | |
| FastGANResolution=1024x1024, Image number=2k2021.01 | 19.01 | — | |
| DMNNFE=52026.03 | 20.93 | — | |
| LD3NFE=32026.03 | 23.86 | — | |
| BaselineResolution=1024x1024, Image number=10k2021.01 | 25.12 | — | |
| NVAEType=VAE2022.04 | 26.02 | — | |
| GITSNFE=32026.03 | 26.41 | — | |
| AMEDNFE=32026.03 | 26.87 | — | |
| AMEDNFE=42026.03 | 26.89 | — | |
| BaselineResolution=1024x1024, Image number=5k2021.01 | 27.86 | — | |
| best heu.NFE=42026.03 | 29.35 | — | |
| DMNNFE=42026.03 | 31.3 | — | |
| VDVAEType=VAE2022.04 | 33.5 | — | |
| BaselineResolution=1024x1024, Image number=2k2021.01 | 36.45 | — | |
| GPE (Latent Diffusion)Representation Space=Latent, Encoder=GPE, Architecture=U-Net, Framework=EDM2026.04 | 48.16 | — | |
| PSeg2023.05 | 62.44 | — | |
| VAE (Latent Diffusion)Representation Space=Latent, Encoder=VAE, Architecture=U-Net, Framework=EDM2026.04 | 62.51 | — | |
| best heu.NFE=32026.03 | 72.29 | — | |
| DMNNFE=32026.03 | 178.09 | — |