Class-Incremental Learning on ImageNet-R (10 sessions, 20-way)
83.12Task Accuracy (AT)FeCAM
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| FeCAMBackbone=DINOv2 ViT-B/142026.05 | 83.12 | 87.78 | 4.4 | |
| COMPOSEBackbone=DINOv2 ViT-B/142026.05 | 82.3 | 86.13 | 5.97 | |
| F-OALBackbone=DINOv2 ViT-B/142026.05 | 81.7 | 87.04 | 6.45 | |
| RanPACBackbone=DINOv2 ViT-B/142026.05 | 81.6 | 86.31 | 8.18 | |
| COMPOSE-CTBackbone=DINOv2 ViT-B/142026.05 | 80.57 | 84.92 | 6.13 |