Image Classification on ImageNet 10% labels subset ILSVRC-2012 (val)
89.9Top-5 AccuracySWAV
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SWAVBackbone=ResNet-50, Evaluation protocol=Semi-supervised fine-tuning2021.03 | 89.9 | 70.2 | |
| BARLOW TWINSBackbone=ResNet-50, Evaluation protocol=Semi-supervised fine-tuning2021.03 | 89.3 | 69.7 | |
| BYOLBackbone=ResNet-50, Evaluation protocol=Semi-supervised fine-tuning2021.03 | 89 | 68.8 | |
| SIMCLRBackbone=ResNet-50, Evaluation protocol=Semi-supervised fine-tuning2021.03 | 87.8 | 65.6 | |
| S4L-RotationBackbone=ResNet50v22019.05 | 83.82 | — | |
| PIRLBackbone=ResNet-50, Evaluation protocol=Semi-supervised fine-tuning2021.03 | 83.8 | — | |
| S4L-ExemplarBackbone=ResNet50v22019.05 | 83.72 | — | |
| VAT + Entropy MinimizationBackbone=ResNet50v22019.05 | 83.39 | — | |
| VATBackbone=ResNet50v22019.05 | 82.78 | — | |
| PseudolabelsBackbone=ResNet50v22019.05 | 82.41 | — | |
| Self-sup. ExemplarBackbone=ResNet50v2, Evaluation Protocol=Fine-tune2019.05 | 81.01 | — | |
| Supervised BaselineBackbone=ResNet50v22019.05 | 80.43 | — | |
| SupervisedBackbone=ResNet-50, Evaluation protocol=Fine-tuning2021.03 | 80.4 | 56.4 | |
| Self-sup. RotationBackbone=ResNet50v2, Evaluation Protocol=Fine-tune2019.05 | 78.53 | — | |
| Self-sup. RotationBackbone=ResNet50v2, Evaluation Protocol=Linear2019.05 | 39.75 | — | |
| Self-sup. ExemplarBackbone=ResNet50v2, Evaluation Protocol=Linear2019.05 | 32.32 | — |