Conditional Image Generation on CIFAR-10 (test)
1.36Fréchet Inception DistanceStyleSAN-XL
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| StyleSAN-XLFamily=Diffusion & GAN, NFE=12025.09 | 1.36 | — | — | |
| SiD2AFamily=Diffusion & GAN, NFE=1, alpha_SiD=1.22025.09 | 1.39 | — | — | |
| SiD2AFamily=Diffusion & GAN, NFE=1, alpha_SiD=1.02025.09 | 1.4 | — | — | |
| GDD-INFE=12024.05 | 1.44 | — | — | |
| GDD-IFamily=Diffusion & GAN, NFE=12025.09 | 1.44 | — | — | |
| SiDAFamily=Diffusion & GAN, NFE=1, alpha_SiD=1.02025.09 | 1.44 | — | — | |
| GDDNFE=12024.05 | 1.58 | — | — | |
| CTMNFE=22024.05 | 1.63 | — | — | |
| EDM-G++Space=Data, NFE=352022.11 | 1.64 | — | — | |
| SiDNFE=1, alpha=1.22024.05 | 1.71 | — | — | |
| SiDFamily=Diffusion & GAN, NFE=1, alpha_SiD=1.22025.09 | 1.71 | — | — | |
| CTMNFE=12024.05 | 1.73 | — | — | |
| CTMFamily=Diffusion & GAN, NFE=12025.09 | 1.73 | — | — | |
| EDMSpace=Data, NFE=35, Seed=manual2022.11 | 1.79 | — | — | |
| EDMNFE=352024.05 | 1.79 | — | — | |
| VP-EDMFamily=Diffusion & GAN, NFE=352025.09 | 1.79 | — | — | |
| EDMSpace=Data, NFE=39, Seed=random2022.11 | 1.82 | — | — | |
| StyleGAN-XLFamily=Diffusion & GAN, NFE=12025.09 | 1.85 | — | — | |
| RealUID + FTFamily=Flow-based, NFE=12025.09 | 1.87 | — | — | |
| SiDNFE=1, alpha=1.02024.05 | 1.93 | — | — | |
| SiDFamily=Diffusion & GAN, NFE=1, alpha_SiD=1.02025.09 | 1.93 | — | — | |
| Score SDENFE=20002024.05 | 2.2 | — | — | |
| NCSN++-GSpace=Data, NFE=20002022.11 | 2.25 | — | — | |
| StyleGAN2+ADA+Tune+DIFamily=Diffusion & GAN, NFE=12025.09 | 2.27 | — | — | |
| StyleGAN2-ADAMode=Conditional2020.11 | 2.42 | 10.14 | — | |
| StyleGAN2-ADANFE=12023.12 | 2.42 | 10.14 | — | |
| StyleGAN2-ADANFE=12024.05 | 2.42 | — | — | |
| StyleGAN2+ADA+TuneFamily=Diffusion & GAN, NFE=12025.09 | 2.42 | — | — | |
| FGMFamily=Flow-based, NFE=12025.09 | 2.58 | — | — | |
| DPM-Solve-3NFE=482024.05 | 2.65 | — | — | |
| DMDFamily=Diffusion & GAN, NFE=12025.09 | 2.66 | — | — | |
| StyleGAN2 + ADA (v1)2020.06 | 2.67 | 10.06 | — | |
| StyleGAN2+ADAFamily=Diffusion & GAN, NFE=12025.09 | 3.49 | — | — | |
| CUOTM+SD†NFE=12026.03 | 3.71 | 8.83 | 13.44 | |
| DMD (w.o. KL)Family=Diffusion & GAN, NFE=12025.09 | 3.82 | — | — | |
| Diff BigGAN + BIGRoCConditioning=Conditional, BIGRoC Variant=GT2021.08 | 3.95 | 9.61 | — | |
| OT Bayesian FlowNFE=1002026.03 | 4.1 | 8.81 | 14.98 | |
| Diff-InstructFamily=Diffusion & GAN, NFE=12025.09 | 4.19 | — | — | |
| Diff BigGANConditioning=Conditional2021.08 | 4.37 | 9.48 | — | |
| CUOTM†NFE=12026.03 | 5.3 | 8.79 | 15.55 | |
| DMD (w.o. reg.)Family=Diffusion & GAN, NFE=12025.09 | 5.58 | — | — | |
| GET-BaseNFE=12023.12 | 6.25 | 9.4 | — | |
| GET-BaseFamily=Diffusion & GAN, NFE=12025.09 | 6.25 | — | — | |
| BigGAN + BIGRoCConditioning=Conditional, BIGRoC Variant=GT2021.08 | 6.79 | 9.47 | — | |
| aw-BigGANType=Conditional2020.12 | 7.03 | 9.52 | — | |
| Guided DistillationNFE=1, Classifier guidance level (w)=0.32023.12 | 7.34 | 8.9 | — | |
| BigGANConditioning=Conditional2021.08 | 7.45 | 9.38 | — | |
| MHingeType=Conditional2020.12 | 7.5 | 9.58 | — | |
| BigGAN-SaBNNormalization Layer=SaBN2021.02 | 8.03 | 9.01 | — | |
| aw-SNGANType=Conditional2020.12 | 8.11 | 9.03 | — | |
| cGAN-PD + BIGRoCConditioning=Conditional, BIGRoC Variant=GT2021.08 | 8.33 | 8.76 | — | |
| Guided DistillationNFE=1, Classifier guidance level (w)=02023.12 | 8.34 | 8.63 | — | |
| BigGAN+TuneFamily=Diffusion & GAN, NFE=12025.09 | 8.47 | — | — | |
| BigGANNormalization Layer=CCBN2021.02 | 8.57 | 8.91 | — | |
| Guided DistillationNFE=1, Classifier guidance level (w)=12023.12 | 8.62 | 9.21 | — | |
| cGAN + BIGRoCConditioning=Conditional, BIGRoC Variant=GT2021.08 | 8.89 | 8.57 | — | |
| SNGAN-SaBNNormalization Layer=SaBN2021.02 | 8.97 | 8.89 | — | |
| AutoGAN-top1-SaBNNormalization Layer=SaBN2021.02 | 9.11 | 8.72 | — | |
| NDA (Jigsaw)Base Model=BigGAN, NDA Transformation=Jigsaw, lambda=0.252021.02 | 9.42 | — | — | |
| NDA (Stitching)Base Model=BigGAN, NDA Transformation=Stitching, lambda=0.252021.02 | 9.47 | — | — | |
| CR+LT-BigGANBackbone=BigGAN, Regularization=CR+LT, Setting=Conditional2020.10 | 9.8 | — | — | |
| SNGANNormalization Layer=CCBN2021.02 | 10.18 | 8.76 | — | |
| NDA (Cutmix)Base Model=BigGAN, NDA Transformation=Cutmix, lambda=0.252021.02 | 10.3 | — | — | |
| AutoGAN-top1Normalization Layer=CCBN2021.02 | 10.51 | 8.43 | — | |
| NDA (Cutout)Base Model=BigGAN, NDA Transformation=Cutout, lambda=0.252021.02 | 10.52 | — | — | |
| LT-BigGANBackbone=BigGAN, Setting=Conditional2020.10 | 11.01 | — | — | |
| cGAN-PDConditioning=Conditional2021.08 | 11.1 | 8.54 | — | |
| CR-BigGANBackbone=BigGAN, Regularization=CR, Setting=Conditional2020.10 | 11.48 | — | — | |
| CR-BigGANBase Model=BigGAN, Regularization=Consistency Regularization2021.02 | 11.48 | — | — | |
| BigGANAugmentation=None, Base Model=BigGAN2021.02 | 11.51 | — | — | |
| Guided DistillationNFE=1, Classifier guidance level (w)=22023.12 | 13.23 | 9.23 | — | |
| NDA (Mixup)Base Model=BigGAN, NDA Transformation=Mixup, lambda=0.252021.02 | 13.87 | — | — | |
| BigGANBackbone=BigGAN, Setting=Conditional2020.10 | 14.73 | — | — | |
| BigGAN2020.06 | 14.73 | 9.22 | — | |
| BigGANMode=Conditional2020.11 | 14.73 | 9.22 | — | |
| BigGANType=Conditional2020.12 | 14.73 | 9.22 | — | |
| BigGANNFE=12023.12 | 14.73 | 9.22 | — | |
| BigGANNFE=12024.05 | 14.73 | — | — | |
| BigGANFamily=Diffusion & GAN, NFE=12025.09 | 14.73 | — | — | |
| CGAN-PDGAN Architecture=GAN in [34], Supervision=fully labeled data2019.12 | 17.5 | 8.62 | — | |
| cGANType=Conditional2020.12 | 17.5 | 8.62 | — | |
| SNGANType=Conditional2020.12 | 17.5 | 8.6 | — | |
| Triple-GAN-V2GAN Architecture=GAN in [34], Supervision=partially labeled (4,000 labels)2019.12 | 17.9 | 8.51 | — | |
| ACGANGAN Architecture=ResNet GAN, Supervision=partially labeled (4,000 labels)2019.12 | 19.7 | 8.22 | — | |
| SNGANGAN Architecture=GAN in [34], Supervision=fully unlabeled data2019.12 | 21.7 | 8.22 | — | |
| WGAN-GPGAN Architecture=ResNet GAN, Supervision=fully unlabeled data2019.12 | 24.8 | 7.86 | — | |
| cGANConditioning=Conditional2021.08 | 29.26 | 6.95 | — | |
| COTM†NFE=12026.03 | 33.04 | 6.58 | 65.66 | |
| DCGANGAN Architecture=ResNet GAN, Supervision=fully unlabeled data2019.12 | 36.9 | 6.16 | — | |
| EBM2020.06 | 37.9 | 8.3 | — | |
| JEM2020.06 | 38.4 | 8.76 | — | |
| CGAN-PDGAN Architecture=GAN in [34], Supervision=small subset of fully labeled data (4,000 samples), Source=our code base2019.12 | 42.3 | 6.03 | — | |
| DCGANGAN Architecture=ResNet GAN, Supervision=partially labeled (4,000 labels)2019.12 | — | 6.58 | — | |
| Improved-GAN-FMGAN Architecture=GAN in [26], Supervision=partially labeled (4,000 labels)2019.12 | — | 3.87 | — | |
| Triple-GAN-V1GAN Architecture=GAN in [26], Supervision=partially labeled (4,000 labels)2019.12 | — | 5.08 | — | |
| WGAN-GPGAN Architecture=ResNet GAN, Supervision=fully labeled data2019.12 | — | 8.42 | — |