Image Classification on ImageNet (test) (1%/10% Label Subset Accuracy)
77.9Top-1 Acc (10% Labels)CPC v2
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| CPC v2Architecture=ResNet-161, Parameters=305M2020.06 | 77.9 | — | — | 91.2 | |
| BYOLArchitecture=ResNet-200 (2x), Parameters=250M2020.06 | 77.7 | 71.2 | 89.5 | 93.7 | |
| BYOLArchitecture=ResNet-50 (4x), Parameters=375M2020.06 | 75.7 | 69.1 | 87.9 | 92.5 | |
| SimCLRArchitecture=ResNet-50 (4x), Parameters=375M2020.06 | 74.4 | 63 | 85.8 | 92.6 | |
| BYOLArchitecture=ResNet-50 (2x), Parameters=94M2020.06 | 73.5 | 62.2 | 84.1 | 91.7 | |
| SimCLRArchitecture=ResNet-50 (2x), Parameters=94M2020.06 | 71.7 | 58.5 | 83 | 91.2 | |
| BYOLBackbone=ResNet-50, Evaluation Protocol=Semi-supervised fine-tuning2020.06 | 68.8 | 53.2 | 78.4 | 89 | |
| SimCLRBackbone=ResNet-50, Evaluation Protocol=Semi-supervised fine-tuning2020.06 | 65.6 | 48.3 | 75.5 | 87.8 | |
| PIRLBackbone=ResNet-50, Evaluation Protocol=Semi-supervised fine-tuning2020.06 | 57.2 | — | — | 83.8 | |
| SupervisedBackbone=ResNet-50, Evaluation Protocol=Semi-supervised fine-tuning2020.06 | 56.4 | 25.4 | 48.4 | 80.4 | |
| InstDiscBackbone=ResNet-50, Evaluation Protocol=Semi-supervised fine-tuning2020.06 | 39.2 | — | — | 77.4 |