Class-conditional image generation on ImageNet (test)
6.04FIDStackformer
Evaluation results
| Method | Links | ||
|---|---|---|---|
| StackformerModel Type=Autoregressive, #Params=651M2023.05 | 6.04 | 172.6 | |
| MaskGITModel Type=Bidirection, #Params=227M2023.05 | 6.18 | 182.1 | |
| BigGAN-deepModel Type=GAN, #Params=112M2023.05 | 6.84 | 203.6 | |
| Mo-VQGANModel Type=Autoregressive, #Params=383M2023.05 | 7.13 | 138.3 | |
| BigGANModel Type=GAN, #Params=164M2023.05 | 7.53 | 168.6 | |
| ADMModel Type=Diffusion, #Params=554M2023.05 | 10.94 | 101 | |
| RQ-TransformerModel Type=Autoregressive, #Params=821M2023.05 | 13.11 | 104.3 | |
| VQ-GAN†Model Type=Autoregressive, #Params=679M, Note=trained with same setup as current paper2023.05 | 17.03 | 76.85 | |
| DCTModel Type=Autoregressive, #Params=738M2023.05 | 36.5 | — |