Long-tailed Image Recognition on CIFAR-100 Long-Tailed (test)
67.4Accuracy (Old)BaCon-S
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| BaCon-SBackbone=ViT-B/16, Pre-trained=DINO, Fine-tuning protocol=fine-tune last block, pseudo-labeling branch=SimGCD-style2023.10 | 67.4 | 66.5 | 67.2 | |
| BaCon-OBackbone=ViT-B/16, Pre-trained=DINO, Fine-tuning protocol=fine-tune last block, pseudo-labeling branch=ORCA-style2023.10 | 66.5 | 69.6 | 67.1 | |
| GCDBackbone=ViT-B/16, Pre-trained=DINO, Fine-tuning protocol=fine-tune last block2023.10 | 65.5 | 49 | 62.2 | |
| OpenConBackbone=ViT-B/16, Pre-trained=DINO, Fine-tuning protocol=fine-tune last block2023.10 | 64.2 | 40.9 | 59.6 | |
| SimGCDBackbone=ViT-B/16, Pre-trained=DINO, Fine-tuning protocol=fine-tune last block2023.10 | 59.8 | 24.6 | 52.8 | |
| TRSSLBackbone=ViT-B/16, Pre-trained=DINO, Fine-tuning protocol=fine-tune last block, Adapted=true2023.10 | 58.7 | 35.8 | 54.1 | |
| ORCABackbone=ViT-B/16, Pre-trained=DINO, Fine-tuning protocol=fine-tune last block2023.10 | 55 | 30.8 | 50.1 | |
| DARPBackbone=ViT-B/16, Pre-trained=DINO, Fine-tuning protocol=fine-tune last block2023.10 | 54.5 | 24 | 48.4 | |
| ABCBackbone=ViT-B/16, Pre-trained=DINO, Fine-tuning protocol=fine-tune last block2023.10 | 48.5 | 20.8 | 43 |