Image Classification on iNaturalist 2021 (val)
85.1AccuracyViT-B w/ AGLU
Evaluation results
| Method | Links | |
|---|---|---|
| ViT-B w/ AGLUBackbone=ViT-B, AGLU=true, Crop Size=2242024.07 | 85.1 | |
| ViT-BBackbone=ViT-B, Crop Size=2242024.07 | 84.7 | |
| SE-ResNet50 w/ APA* + AGLUBackbone=ResNet-50, SE=true, APA=true, AGLU=true, Crop Size=2242024.07 | 83.9 | |
| LoSABackbone=ViT-g (1B)2024.02 | 83.8 | |
| LoRABackbone=ViT-g (1B)2024.02 | 83.3 | |
| Full fine-tuningBackbone=ViT-g (1B)2024.02 | 82.5 | |
| SE-ResNet50Backbone=ResNet-50, SE=true, Crop Size=2242024.07 | 82.1 | |
| Finetune only attentionBackbone=ViT-g (1B)2024.02 | 81.9 | |
| Finetune only mlpBackbone=ViT-g (1B)2024.02 | 81.7 | |
| Finetune last 2 layersBackbone=ViT-g (1B)2024.02 | 79.8 | |
| Finetune last layerBackbone=ViT-g (1B)2024.02 | 79.2 | |
| Prompt TuningBackbone=ViT-g (1B)2024.02 | 76.8 | |
| BitfitBackbone=ViT-g (1B)2024.02 | 76.4 | |
| LSTBackbone=ViT-g (1B)2024.02 | 76 | |
| Linear probingBackbone=ViT-g (1B)2024.02 | 70.1 |