Image Classification on CIFAR-10-LT λ=200 (test)
23.11Error RateMetaSAug with cross-entropy loss
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| MetaSAug with cross-entropy lossBackbone=ResNet-322021.12 | 23.11 | — | — | — | |
| RISDABackbone=ResNet-322021.12 | 26 | — | — | — | |
| Meta-class-weight with cross-entropy lossBackbone=ResNet-322021.12 | 29.34 | — | — | — | |
| Class-balanced CEBackbone=ResNet-322021.12 | 31.23 | — | — | — | |
| Meta-weight netBackbone=ResNet-322021.12 | 32.8 | — | — | — | |
| L2RWBackbone=ResNet-322021.12 | 33.75 | — | — | — | |
| Class-balanced fine-tuningBackbone=ResNet-322021.12 | 33.76 | — | — | — | |
| Cross-EntropyBackbone=ResNet-322021.12 | 34.13 | — | — | — | |
| CEloss=Cross-Entropy2023.03 | — | 37.8 | 42.4 | 64 | |
| CE+ALMloss=Cross-Entropy + ALM2023.03 | — | 36.1 | 39.9 | 65.1 | |
| CE+RankRegloss=Cross-Entropy + RankReg2023.03 | — | 36.7 | 37.8 | 65 |