Image Classification on mini-ImageNet-Blue (80% noise)
51.8AccuracyMixup
Evaluation results
| Method | Links | |
|---|---|---|
| MixupBackbone=ResNet-18, Training Protocol=Oracle: Clean Subset, Data Augmentation=Mixup2022.02 | 51.8 | |
| StandardBackbone=ResNet-18, Training Protocol=Oracle: Clean Subset2022.02 | 50.6 | |
| StandardBackbone=ResNet-18, Training Protocol=Baseline2022.02 | 13.4 | |
| BootstrapBackbone=ResNet-18, Training Protocol=Baseline2022.02 | 13.4 | |
| NCRBackbone=ResNet-18, Training Protocol=NCR2022.02 | 11.5 | |
| Label smoothingBackbone=ResNet-18, Training Protocol=Baseline2022.02 | 9.1 | |
| MixupBackbone=ResNet-18, Training Protocol=Baseline, Data Augmentation=Mixup2022.02 | 7.8 | |
| Bootstrap + MixupBackbone=ResNet-18, Training Protocol=Baseline, Data Augmentation=Mixup2022.02 | 7.6 | |
| NCR + MixupBackbone=ResNet-18, Training Protocol=NCR, Data Augmentation=Mixup2022.02 | 7.1 | |
| Label smoothing + MixupBackbone=ResNet-18, Training Protocol=Baseline, Data Augmentation=Mixup2022.02 | 6.8 |