Long-tailed Visual Recognition on iNaturalist 2017 (test)
75.9AccuracyMAE
Evaluation results
| Method | Links | |
|---|---|---|
| MAEtraining_mode=fine-tune, pre-training=ImageNet-1K2024.11 | 75.9 | |
| CBDENSBackbone=ResNet-1012021.04 | 71.3 | |
| CBDENSBackbone=ResNet-502021.04 | 69.3 | |
| Attention Distillationpre-training=ImageNet-1K2024.11 | 69.3 | |
| Attention Copypre-training=ImageNet-1K2024.11 | 69.1 | |
| CBDBackbone=ResNet-1012021.04 | 66.5 | |
| Feature Aug.Backbone=ResNet-1012021.04 | 65.9 | |
| BBNBackbone=ResNet-502021.04 | 65.8 | |
| CRTBackbone=ResNet-1012021.04 | 65.2 | |
| CBDBackbone=ResNet-502021.04 | 64.6 | |
| CRTBackbone=ResNet-502021.04 | 63.9 | |
| Feature Aug.Backbone=ResNet-502021.04 | 62 | |
| CBBackbone=ResNet-1012021.04 | 60.9 | |
| Rethinking CBBackbone=ResNet-502021.04 | 59.4 | |
| CBBackbone=ResNet-502021.04 | 58.1 | |
| MAEtraining_mode=scratch2024.11 | 49.7 |