Classification on CelebA WILDS
95.1Average AccuracyERM
Evaluation results
| Method | Links | ||
|---|---|---|---|
| ERMBackbone=ResNet-50, Pre-trained=ImageNet, Selection Criteria=Worst group accuracy2024.11 | 95.1 | 62.6 | |
| IRMBackbone=ResNet-50, Pre-trained=ImageNet, Selection Criteria=Worst group accuracy2024.11 | 94.7 | 63 | |
| CeGDROBackbone=ResNet-50, Pre-trained=ImageNet, Selection Criteria=Worst group accuracy2024.11 | 91.9 | 89.8 | |
| CeGDRO - EFBackbone=ResNet-50, Pre-trained=ImageNet, Selection Criteria=Worst group accuracy2024.11 | 91.7 | 89.4 | |
| GroupDROBackbone=ResNet-50, Pre-trained=ImageNet, Selection Criteria=Worst group accuracy2024.11 | 91.4 | 89 | |
| GroupDRO + SCBackbone=ResNet-50, Pre-trained=ImageNet, Selection Criteria=Worst group accuracy2024.11 | 86.7 | 0 |