Image Classification on mini-ImageNet Blue (20% noise)
70.4AccuracyStandard
Evaluation results
| Method | Links | |
|---|---|---|
| StandardBackbone=ResNet-18, Training Protocol=Oracle: Clean Subset2022.02 | 70.4 | |
| MixupBackbone=ResNet-18, Training Protocol=Oracle: Clean Subset, Data Augmentation=Mixup2022.02 | 70.3 | |
| NCR + MixupBackbone=ResNet-18, Training Protocol=NCR, Data Augmentation=Mixup2022.02 | 68.3 | |
| Label smoothing + MixupBackbone=ResNet-18, Training Protocol=Baseline, Data Augmentation=Mixup2022.02 | 67.9 | |
| NCRBackbone=ResNet-18, Training Protocol=NCR2022.02 | 67.8 | |
| Label smoothingBackbone=ResNet-18, Training Protocol=Baseline2022.02 | 67.7 | |
| MixupBackbone=ResNet-18, Training Protocol=Baseline, Data Augmentation=Mixup2022.02 | 66.5 | |
| BootstrapBackbone=ResNet-18, Training Protocol=Baseline2022.02 | 66.5 | |
| Bootstrap + MixupBackbone=ResNet-18, Training Protocol=Baseline, Data Augmentation=Mixup2022.02 | 64.6 | |
| StandardBackbone=ResNet-18, Training Protocol=Baseline2022.02 | 63.4 |