Image Classification on ImageNet 10% label fraction
95.5Top-5 AccSimCLRv2 self-distilled
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SimCLRv2 self-distilledArchitecture=ResNet-152 (3x+SK)2020.06 | 95.5 | 80.9 | |
| SimCLRv2 distilledArchitecture=ResNet-50 (2x+SK)2020.06 | 95 | 80.2 | |
| BYOLArchitecture=ResNet-200 (2x)2020.06 | 93.7 | 77.7 | |
| SimCLRv2 distilledArchitecture=ResNet-502020.06 | 93.4 | 77.5 | |
| SimCLRArchitecture=ResNet-50 (4x), Evaluation Protocol=Fine-tuning, Data Augmentation=Broader composition2020.02 | 92.8 | 74.8 | |
| SimCLRArchitecture=ResNet-50 (4x)2020.06 | 92.6 | 74.4 | |
| SimCLRArchitecture=ResNet-50 (2x), Evaluation Protocol=Fine-tuning, Data Augmentation=Broader composition2020.02 | 91.2 | 71.8 | |
| S4L (Rot+VAT+Entropy Min.)Architecture=ResNet-50 (4x)2020.06 | 91.2 | 73.2 | |
| CowMixArchitecture=ResNet-1522020.06 | 91.2 | 73.9 | |
| CPC v2Architecture=ResNet-161(*)2020.06 | 91.2 | 73.1 | |
| SimCLRArchitecture=ResNet-50 (2x)2020.06 | 91.2 | 71.7 | |
| Mean teacherArchitecture=ResNeXt-1522020.06 | 90.9 | — | |
| FixMatch (w. RandAug)Architecture=ResNet-502020.06 | 89.1 | 71.5 | |
| BYOLArchitecture=ResNet-502020.06 | 89 | 68.8 | |
| UDA (w. RandAug)Architecture=ResNet-502020.06 | 88.5 | 68.8 | |
| SimCLRArchitecture=ResNet-50, Evaluation Protocol=Fine-tuning, Data Augmentation=Broader composition2020.02 | 88.1 | 66.1 | |
| SimCLRArchitecture=ResNet-502020.06 | 87.8 | 65.6 | |
| PIRLArchitecture=ResNet-502020.06 | 83.8 | — | |
| VAT+Entropy Min.Architecture=ResNet-502020.06 | 83.4 | — | |
| Pseudo-labelArchitecture=ResNet-502020.06 | 82.4 | — | |
| Supervised baselineArchitecture=ResNet-502020.06 | 80.4 | 56.4 | |
| BigBiGANArchitecture=RevNet-50 (4x)2020.06 | 78.8 | — | |
| InstDiscArchitecture=ResNet-502020.06 | 77.4 | — | |
| MPL (w. RandAug)Architecture=ResNet-502020.06 | — | 73.8 |