Image Classification on CIFAR-10 Imbalance Ratio 100 Heteroskedastic Noise (val)
42.4Top-1 Acc (Noisy & Rare)HAR
Evaluation results
| Method | Links | ||
|---|---|---|---|
| HARBackbone=ResNet-322020.06 | 42.4 | 94 | |
| Unif-reg (optimal)Backbone=ResNet-322020.06 | 36.7 | 92.4 | |
| MW-NetBackbone=ResNet-322020.06 | 35.6 | 92.3 | |
| L2RWBackbone=ResNet-322020.06 | 31.2 | 89.7 | |
| MentorNetBackbone=ResNet-322020.06 | 28.3 | 90.2 | |
| INCVBackbone=ResNet-322020.06 | 25.8 | 93.8 | |
| ERMBackbone=ResNet-322020.06 | 18.9 | 94.2 | |
| Co-teachingBackbone=ResNet-322020.06 | 15.4 | 86.4 | |
| GCEBackbone=ResNet-322020.06 | 14.1 | 91.7 | |
| DMIBackbone=ResNet-322020.06 | 14 | 91.8 |