Long-tailed Recognition on Long-Tailed CIFAR-100 (test) (Accuracy by Imbalance Factor)
46.03Top-1 Acc (IF 100)MW-Net with LDAM loss + FaMUS
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| MW-Net with LDAM loss + FaMUSBackbone=ResNet-32, Extra Clean Data=true2021.04 | 46.03 | 49.93 | 55.95 | 59.03 | |
| MW-Net with CE loss + FaMUSBackbone=ResNet-32, Extra Clean Data=true2021.04 | 45.6 | 49.56 | 56.22 | 60.42 | |
| LDAM-DRWBackbone=ResNet-32, Extra Clean Data=false2021.04 | 44.7 | — | — | 59.59 | |
| Improved L2R with LDAMBackbone=ResNet-32, Extra Clean Data=true2021.04 | 44.08 | 49.16 | 52.38 | 58 | |
| Improved L2R with CE lossBackbone=ResNet-32, Extra Clean Data=true2021.04 | 43.35 | 48.53 | 55.62 | 59.58 | |
| BBNBackbone=ResNet-32, Extra Clean Data=false2021.04 | 42.56 | 47.02 | — | 59.12 | |
| MW-Net with CE lossBackbone=ResNet-32, Extra Clean Data=true2021.04 | 42.09 | 46.74 | 54.37 | 58.46 | |
| L2R+ with CE lossBackbone=ResNet-32, Extra Clean Data=false2021.04 | 40.23 | 44.44 | 51.64 | 53.73 | |
| CB FocalBackbone=ResNet-32, Extra Clean Data=false2021.04 | 39.6 | 45.32 | 52.59 | 57.99 | |
| Focal LossBackbone=ResNet-32, Extra Clean Data=false2021.04 | 38.41 | 44.32 | 51.95 | 55.78 | |
| CE lossBackbone=ResNet-32, Extra Clean Data=false2021.04 | 38.32 | 43.85 | 51.14 | 55.71 |