Image Classification on CIFAR-LT-100 Imbalance Factor 10 (test)
68.67Top-1 AccuracyWeight Balancing
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Weight Balancingregularization=+ WD & Max2022.03 | 68.67 | — | |
| Weight Balancingregularization=+ WD2022.03 | 67.96 | — | |
| Weight Balancingregularization=+ tau-norm2022.03 | 67.79 | — | |
| Weight Balancingregularization=+ L2norm2022.03 | 67.16 | — | |
| Weight Balancingregularization=+ Max2022.03 | 67.1 | — | |
| ResComBackbone=ResNet-322022.03 | 66.1 | — | |
| Weight Balancingregularization=WD2022.03 | 66.03 | — | |
| PaCo2022.03 | 64.2 | — | |
| PaCoBackbone=ResNet-322022.03 | 64.2 | — | |
| tau-norm2022.03 | 63.8 | — | |
| DRO-LT2022.03 | 63.41 | — | |
| MiSLASBackbone=ResNet-322022.03 | 63.2 | — | |
| Balanced SoftmaxBackbone=ResNet-322022.03 | 63 | — | |
| Hybrid-PSCBackbone=ResNet-322022.03 | 62.4 | — | |
| SSD2022.03 | 62.3 | — | |
| DiVE2022.03 | 62 | — | |
| DiVEImbalance Factor=102021.03 | 62 | — | |
| RIDEBackbone=ResNet-32, experts=32022.03 | 61.8 | — | |
| LADEBackbone=ResNet-322022.03 | 61.7 | — | |
| ResLTBackbone=ResNet-322022.03 | 60.8 | — | |
| De-confound2022.03 | 59.6 | — | |
| Causal NormBackbone=ResNet-322022.03 | 59.6 | — | |
| TDEImbalance Factor=102021.03 | 59.6 | — | |
| Meta-learningImbalance Factor=102021.03 | 59.59 | — | |
| KD2022.03 | 59.22 | — | |
| BBN2022.03 | 59.12 | — | |
| BBNImbalance Factor=102021.03 | 59.12 | — | |
| BBNBackbone=ResNet-322022.03 | 59.1 | — | |
| LDAM+SSP2022.03 | 58.91 | — | |
| LDAM-DRW+SSPImbalance Factor=102021.03 | 58.91 | — | |
| LDAM-DRW2022.03 | 58.71 | — | |
| LDAM-DRWImbalance Factor=102021.03 | 58.71 | — | |
| BSCEImbalance Factor=102021.03 | 58.38 | — | |
| CE+CB2022.03 | 57.99 | — | |
| Focal+CB2022.03 | 57.99 | — | |
| LogitAjust2022.03 | 57.74 | — | |
| Weight Balancingregularization=naive2022.03 | 57.31 | — | |
| CEImbalance Factor=102021.03 | 56.51 | — | |
| Focal2022.03 | 55.78 | — | |
| Focal+Imbalance Factor=102021.03 | 55.78 | — | |
| CE2022.03 | 55.71 | — | |
| BBNBackbone=ResNet-32, Imbalance factor=102021.03 | — | 40.88 | |
| Class-balanced cross-entropy lossBackbone=ResNet-32, Imbalance factor=102021.03 | — | 42.43 | |
| Class-balanced fine-tuningBackbone=ResNet-32, Imbalance factor=102021.03 | — | 42.43 | |
| Class-balanced focal lossBackbone=ResNet-32, Imbalance factor=102021.03 | — | 42.01 | |
| Cross-entropy trainingBackbone=ResNet-32, Imbalance factor=102021.03 | — | 44.27 | |
| Focal lossBackbone=ResNet-32, Imbalance factor=102021.03 | — | 44.22 | |
| L2RWBackbone=ResNet-32, Imbalance factor=102021.03 | — | 47.88 | |
| LDAM lossBackbone=ResNet-32, Imbalance factor=102021.03 | — | 42.71 | |
| LDAM-DRWBackbone=ResNet-32, Imbalance factor=102021.03 | — | 41.22 | |
| Meta-class-weight with cross-entropy lossBackbone=ResNet-32, Imbalance factor=102021.03 | — | 40.42 | |
| Meta-class-weight with focal lossBackbone=ResNet-32, Imbalance factor=102021.03 | — | 40.41 | |
| Meta-class-weight with LDAM lossBackbone=ResNet-32, Imbalance factor=102021.03 | — | 42 | |
| Meta-weight netBackbone=ResNet-32, Imbalance factor=102021.03 | — | 41.09 | |
| MetaSAug with cross-entropy lossBackbone=ResNet-32, Imbalance factor=102021.03 | — | 38.27 | |
| MetaSAug with focal lossBackbone=ResNet-32, Imbalance factor=102021.03 | — | 38.94 | |
| MetaSAug with LDAM lossBackbone=ResNet-32, Imbalance factor=102021.03 | — | 38.72 | |
| MixupBackbone=ResNet-32, Imbalance factor=102021.03 | — | 41.98 |