Out-of-distribution robustness on 17 Out-of-distribution datasets aggregated (test)
58.2OOD AccuracyResNet-50 (VISSL)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| ResNet-50 (VISSL)Backbone=ResNet-50, Training=VISSL2022.06 | 58.2 | 1 | |
| BC_pruning-fraction-0.3Backbone=ResNet-50, Pruning Method=Best-case (BC), Pruning Fraction=0.32022.06 | 57.8 | 2 | |
| BC_pruning-fraction-0.4Backbone=ResNet-50, Pruning Method=Best-case (BC), Pruning Fraction=0.42022.06 | 57.4 | 3 | |
| BC_pruning-fraction-0.1Backbone=ResNet-50, Pruning Method=Best-case (BC), Pruning Fraction=0.12022.06 | 57.4 | 4 | |
| BC_pruning-fraction-0.2Backbone=ResNet-50, Pruning Method=Best-case (BC), Pruning Fraction=0.22022.06 | 56.5 | 5 | |
| WC_pruning-fraction-0.1Backbone=ResNet-50, Pruning Method=Worst-case (WC), Pruning Fraction=0.12022.06 | 56.5 | 6 | |
| BC_pruning-fraction-0.5Backbone=ResNet-50, Pruning Method=Best-case (BC), Pruning Fraction=0.52022.06 | 56 | 7 | |
| ResNet-50 (torchvision)Backbone=ResNet-50, Training=torchvision2022.06 | 55.9 | 8 | |
| WC_pruning-fraction-0.2Backbone=ResNet-50, Pruning Method=Worst-case (WC), Pruning Fraction=0.22022.06 | 55.6 | 9 | |
| WC_pruning-fraction-0.3Backbone=ResNet-50, Pruning Method=Worst-case (WC), Pruning Fraction=0.32022.06 | 54 | 10 | |
| WC_pruning-fraction-0.4Backbone=ResNet-50, Pruning Method=Worst-case (WC), Pruning Fraction=0.42022.06 | 51.6 | 11 | |
| WC_pruning-fraction-0.5Backbone=ResNet-50, Pruning Method=Worst-case (WC), Pruning Fraction=0.52022.06 | 49.8 | 12 |