Long-tailed Visual Recognition on iNaturalist 2018 (val)
75.3Accuracy (Overall)CBDENS
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| CBDENSBackbone=ResNet-1012021.04 | 75.3 | 77.9 | 76.5 | 73.2 | |
| CBDENSBackbone=ResNet-502021.04 | 73.6 | 75.9 | 74.7 | 71.5 | |
| Logit Adj.Backbone=ResNet-1012021.04 | 70.8 | — | — | — | |
| cRTBackbone=ResNet-1012021.04 | 70.7 | 71.5 | 71.3 | 69.7 | |
| CBDBackbone=ResNet-1012021.04 | 70.5 | 74.2 | 71.5 | 68.3 | |
| ELF(LDAM)Backbone=ResNet-502021.04 | 69.8 | 72.7 | 70.4 | 68.3 | |
| LWSBackbone=ResNet-1012021.04 | 69.7 | 73.9 | 70.4 | 67.8 | |
| BBNBackbone=ResNet-502021.04 | 69.6 | — | — | — | |
| LWSBackbone=ResNet-502021.04 | 69.5 | 71 | 69.8 | 68.8 | |
| Logit Adj.Backbone=ResNet-502021.04 | 68.4 | — | — | — | |
| CBDBackbone=ResNet-502021.04 | 68.4 | 70.5 | 69.5 | 66.5 | |
| cRTBackbone=ResNet-502021.04 | 68.2 | 73.2 | 68.8 | 66.1 | |
| cRT+SSPBackbone=ResNet-502021.04 | 68.1 | — | — | — | |
| ELF(CE)Backbone=ResNet-502021.04 | 66 | 67.4 | 66.3 | 65.1 |