Image Generation on CIFAR-100 (10% data)
12.7Inception ScoreOmniGAN (d=1024) + ADA + CHAIN
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| OmniGAN (d=1024) + ADA + CHAINBackbone=OmniGAN, Massive Augmentation=true2024.03 | 12.7 | — | 9.49 | 14.23 | |
| OmniGAN + LCSA + ADAd'=1024, Augmentation=ADA2021.12 | 12.67 | — | 10.18 | 14.87 | |
| OmniGAN (d=1024) + CHAINBackbone=OmniGAN, Massive Augmentation=false2024.03 | 12.05 | — | 13.12 | 17.87 | |
| OmniGAN + LCSA + DAd'=1024, Augmentation=DA2021.12 | 11.88 | — | 13.68 | 18.44 | |
| OmniGAN + LCSAd'=10242021.12 | 11.85 | — | 13.82 | 18.56 | |
| BigGAN + LCSA + DAd'=256, Augmentation=DA2021.12 | 10.68 | — | 19.38 | 24.21 | |
| BigGAN + LCSA + LeCam + DAd'=256, Loss=LeCam, Augmentation=DA2021.12 | 10.28 | — | 18.24 | 23.12 | |
| OmniGAN + DAd'=1024, Augmentation=DA2021.12 | 10.21 | — | 32.61 | 36.86 | |
| LeCam + DA + CHAINBackbone=BigGAN, Massive Augmentation=true2024.03 | 10.11 | — | 12.69 | 17.49 | |
| BigGAN + LCSAd'=2562021.12 | 10.08 | — | 21.13 | 25.87 | |
| BigGAN (d=256) + CHAINBackbone=BigGAN, Massive Augmentation=false2024.03 | 10.04 | — | 13.13 | 18 | |
| OmniGAN (d=1024) + DABackbone=OmniGAN, Massive Augmentation=true2024.03 | 10.01 | — | 30.68 | 34.94 | |
| BigGAN + LeCam + DAd'=256, Loss=LeCam, Augmentation=DA2021.12 | 9.92 | — | 21.73 | 26.19 | |
| BigGAN + DAd'=256, Augmentation=DA2021.12 | 9.33 | — | 27.01 | 31.32 | |
| LeCam + DABackbone=BigGAN, Massive Augmentation=true2024.03 | 9.17 | — | 22.75 | 27.14 | |
| BigGAN + DigGANBackbone=BigGAN, Regularization=DigGAN2022.11 | 9.06 | 27.61 | — | — | |
| BigGAN + DigGANArchitecture=BigGAN, Regularization=DigGAN2022.11 | 9.06 | 27.61 | — | — | |
| BigGAN + DiffAug + DigGANBackbone=BigGAN, Data Augmentation=DiffAug, Regularization=DigGAN2022.11 | 8.98 | 24.59 | — | — | |
| BigGAN + DiffAug + DigGANArchitecture=BigGAN, Regularization=DigGAN, Data Augmentation=DiffAug2022.11 | 8.98 | 24.59 | — | — | |
| OmniGAN (d=1024) + ADABackbone=OmniGAN, Massive Augmentation=true2024.03 | 8.95 | — | 44.65 | 49.08 | |
| BigGAN + DiffAug + RLCBackbone=BigGAN, Data Augmentation=DiffAug, Regularization=RLC2022.11 | 8.89 | 26.91 | — | — | |
| BigGAN + DiffAug + RLCArchitecture=BigGAN, Regularization=RLC, Data Augmentation=DiffAug2022.11 | 8.89 | 26.91 | — | — | |
| BigGAN (d=256) + DABackbone=BigGAN, Massive Augmentation=true2024.03 | 8.86 | — | 27.22 | 31.8 | |
| OmniGAN + ADAd'=1024, Augmentation=ADA2021.12 | 8.81 | — | 46.74 | 51.03 | |
| BigGAN + DiffAugBackbone=BigGAN, Data Augmentation=DiffAug2022.11 | 8.63 | 33.7 | — | — | |
| BigGAN + DiffAugArchitecture=BigGAN, Data Augmentation=DiffAug2022.11 | 8.63 | 33.7 | — | — | |
| BigGAN (d=256) + LeCamBackbone=BigGAN, Massive Augmentation=false2024.03 | 8.14 | — | 41.51 | 46.43 | |
| BigGAN + LeCamd'=256, Loss=LeCam2021.12 | 8.1 | — | 39.33 | 44.23 | |
| BigGAN + RLCRegularization=RLC2021.04 | 7.76 | — | — | — | |
| BigGAN + RLCBackbone=BigGAN, Regularization=RLC2022.11 | 7.76 | 49.63 | — | — | |
| BigGAN + RLCArchitecture=BigGAN, Regularization=RLC2022.11 | 7.76 | 49.63 | — | — | |
| BigGAN (d=256)Backbone=BigGAN, Massive Augmentation=false2024.03 | 7.58 | — | 50.79 | 55.04 | |
| BigGANd'=2562021.12 | 7.54 | — | 50.89 | 55.15 | |
| LS-GAN2021.04 | 7.02 | — | — | — | |
| RaHinge GAN2021.04 | 6.95 | — | — | — | |
| OmniGAN (d=1024)Backbone=OmniGAN, Massive Augmentation=false2024.03 | 6.91 | — | 60.46 | 64.76 | |
| OmniGANd'=10242021.12 | 6.87 | — | 63.41 | 67.46 | |
| BigGAN2021.04 | 6.74 | — | — | — | |
| Non-saturated GAN2021.04 | 5.96 | — | — | — | |
| BigGANBackbone=BigGAN2022.11 | 5.35 | 73.01 | — | — | |
| BigGANArchitecture=BigGAN2022.11 | 5.35 | 73.01 | — | — | |
| BigGAN2021.04 | — | 66.71 | — | — | |
| BigGAN (d=256) + DigGAN + DABackbone=BigGAN, Massive Augmentation=true2024.03 | — | — | — | 24.59 | |
| BigGAN + RLC2021.04 | — | 49.63 | — | — | |
| DA (Baseline)Backbone=BigGAN2024.03 | — | 35.39 | — | — | |
| DA (Baseline)Backbone=BigGAN2024.03 | — | 35.39 | — | — | |
| DA + KD (CLIP)Backbone=BigGAN2024.03 | — | 33.93 | — | — | |
| DA + VkDBackbone=BigGAN, Whitening=false2024.03 | — | 24.92 | — | — | |
| DA + VkDBackbone=BigGAN, Whitening=true2024.03 | — | 16.87 | — | — | |
| Dense StyleGAN-V2architecture=StyleGAN-V2, mode=Dense, sparsity=0%2021.02 | — | 13.59 | — | — | |
| Dense StyleGAN-V2 + ADAarchitecture=StyleGAN-V2, mode=Dense, augmentation=ADA2021.02 | — | 12.87 | — | — | |
| Dense StyleGAN-V2 + ADA + RLCarchitecture=StyleGAN-V2, mode=Dense, augmentation=ADA + RLC2021.02 | — | 13.01 | — | — | |
| Dense StyleGAN-V2 + DiffAugarchitecture=StyleGAN-V2, mode=Dense, augmentation=DiffAug2021.02 | — | 12.9 | — | — | |
| FitNetsBackbone=BigGAN2024.03 | — | 33.93 | — | — | |
| KD-DLGANBackbone=BigGAN2024.03 | — | 18.03 | — | — | |
| KD-DLGANBackbone=BigGAN2024.03 | — | 18.03 | — | — | |
| Label DistillationBackbone=BigGAN2024.03 | — | 34.14 | — | — | |
| LeCam + DA + KDDLGANBackbone=BigGAN, Massive Augmentation=true2024.03 | — | — | — | 22.4 | |
| LS GAN2021.04 | — | 54.69 | — | — | |
| Non saturated GAN2021.04 | — | 70.5 | — | — | |
| PKDBackbone=BigGAN2024.03 | — | 32.15 | — | — | |
| RaHinge GAN2021.04 | — | 52.72 | — | — | |
| Sparse StyleGAN-V2 Ticketsarchitecture=StyleGAN-V2, mode=Sparse, sparsity=48.80%2021.02 | — | 13.05 | — | — | |
| Sparse StyleGAN-V2 Tickets + ADAarchitecture=StyleGAN-V2, mode=Sparse, sparsity=48.80%, augmentation=ADA2021.02 | — | 12.2 | — | — | |
| Sparse StyleGAN-V2 Tickets + ADA + AdvAugarchitecture=StyleGAN-V2, mode=Sparse, sparsity=48.80%, augmentation=ADA + AdvAug2021.02 | — | 12.11 | — | — | |
| Sparse StyleGAN-V2 Tickets + ADA + RLCarchitecture=StyleGAN-V2, mode=Sparse, sparsity=48.80%, augmentation=ADA + RLC2021.02 | — | 12.48 | — | — | |
| Sparse StyleGAN-V2 Tickets + DiffAugarchitecture=StyleGAN-V2, mode=Sparse, sparsity=48.80%, augmentation=DiffAug2021.02 | — | 12.53 | — | — | |
| SPKDBackbone=BigGAN2024.03 | — | 31.97 | — | — | |
| VkDBackbone=BigGAN, whitening=false2024.03 | — | 24.92 | — | — | |
| VkDBackbone=BigGAN, whitening=true2024.03 | — | 16.87 | — | — |