Image Classification on ImageNet 10% labels (test val)
5.67Top-5 Error RateSup 100%
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Sup 100%Architecture=ResNet-152, Params.=60M2020.03 | 5.67 | 21.33 | |
| SimCLRArchitecture=ResNet-50x4, Params.=375M2020.03 | 7.4 | — | |
| CowMixArchitecture=ResNet-152, Params.=60M2020.03 | 8.76 | 26.06 | |
| S4L Full (MOAM)Architecture=ResNet-50x4, Params.=375M2020.03 | 8.77 | 26.79 | |
| SimCLRArchitecture=ResNet-50x2, Params.=94M2020.03 | 8.8 | — | |
| Mean TeacherArchitecture=ResNeXt-152, Params.=62M2020.03 | 9.11 | — | |
| FixMatchArchitecture=ResNet-50, Params.=24M2020.03 | 10.87 | 28.54 | |
| UDAArchitecture=ResNet-50, Params.=24M2020.03 | 11.2 | 31.22 | |
| SimCLRArchitecture=ResNet-50, Params.=24M2020.03 | 12.2 | — | |
| Sup 10%Architecture=ResNet-152, Params.=60M2020.03 | 22.12 | 42.91 |