Image Classification on CIFAR-10 Imbalance Ratio 10, Heteroskedastic Noise (val)
63.5Top-1 Acc (Noisy&Rare Cls.)HAR
Evaluation results
| Method | Links | ||
|---|---|---|---|
| HARBackbone=ResNet-322020.06 | 63.5 | 94.3 | |
| MW-NetBackbone=ResNet-322020.06 | 55.4 | 91.7 | |
| MentorNetBackbone=ResNet-322020.06 | 54.1 | 90.3 | |
| Unif-reg (optimal)Backbone=ResNet-322020.06 | 53.9 | 92.1 | |
| ERMBackbone=ResNet-322020.06 | 52.9 | 94.4 | |
| INCVBackbone=ResNet-322020.06 | 48.9 | 94 | |
| GCEBackbone=ResNet-322020.06 | 48.2 | 91.6 | |
| DMIBackbone=ResNet-322020.06 | 44.7 | 90.7 | |
| L2RWBackbone=ResNet-322020.06 | 44.3 | 90.1 | |
| Co-teachingBackbone=ResNet-322020.06 | 30.2 | 88.9 |