Long-tailed Image Classification on CIFAR-100 LT (Imbalance ratio 100)
53.41Top-1 AccPrototype Classifier
Evaluation Results
| Method | Links | |
|---|---|---|
| Prototype Classifier2023.02 | 53.41 | |
| WD + WD + Maxweight regularization=WD + WD + Max2023.02 | 53.35 | |
| APA*+AGLU2024.07 | 52.3 | |
| PaCOself-supervised pre-training=true, aggressive data augmentation=true2023.02 | 52 | |
| PaCoTraining setup=RandAugment [12] in 400 epochs2021.07 | 52 | |
| APA*2024.07 | 51.9 | |
| Baseline with SE2024.07 | 50.9 | |
| BALMS2024.07 | 50.8 | |
| CC-SAM2024.07 | 50.8 | |
| Balanced SoftmaxTraining setup=RandAugment [12] in 400 epochs2021.07 | 50.8 | |
| RIDE+CMO+CR2024.07 | 50.7 | |
| RIDE+CMO2024.07 | 50 | |
| TLCnumber of experts=42024.07 | 49.8 | |
| ResLTnumber of experts=32024.07 | 49.7 | |
| ACEnumber of experts=42024.07 | 49.6 | |
| RIDEnumber of experts=42024.07 | 49.4 | |
| RIDEensemble=true2023.02 | 49.1 | |
| AREA2024.07 | 48.9 | |
| GCLBackbone Net=ResNet-322023.05 | 48.71 | |
| NCM2023.02 | 48.67 | |
| MisLASBackbone Net=ResNet-322023.05 | 47.5 | |
| DRO-LT2023.02 | 47.31 | |
| MisLas2024.07 | 47 | |
| MISLAS2021.07 | 47 | |
| Contrastive learningBackbone Net=ResNet-322023.05 | 46.72 | |
| HSC2024.07 | 46.7 | |
| WDweight decay=tuned baseline2023.02 | 46.08 | |
| SSD2024.07 | 46 | |
| DIVE2024.07 | 45.4 | |
| LADE2024.07 | 45.4 | |
| ResLT2021.07 | 45.3 | |
| CE loss + mixup + cRTBackbone Net=ResNet-322023.05 | 45.12 | |
| De-confound-TDEBackbone Net=ResNet-322023.05 | 44.15 | |
| Causal Norm2021.07 | 44.1 | |
| TSC2024.07 | 43.8 | |
| BBN2021.07 | 42.6 | |
| BBNBackbone Net=ResNet-322023.05 | 42.56 | |
| LDAM-DRWBackbone Net=ResNet-322023.05 | 42.04 | |
| LogitAdjust2023.02 | 42.01 | |
| LDAM+DRW2021.07 | 42 | |
| cRT2023.02 | 41.24 | |
| CE loss + mixupBackbone Net=ResNet-322023.05 | 40.01 | |
| LDAM Loss2023.02 | 39.6 | |
| CE lossBackbone Net=ResNet-322023.05 | 38.43 | |
| Focal Loss2023.02 | 38.41 | |
| Focal Loss2021.07 | 38.4 | |
| CE2023.02 | 38.32 |