Class-Incremental Learning on ImageNet-A 10 sessions, 20-way
70.31AT AccuracyCOMPOSE
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| COMPOSEBackbone=DINOv2 ViT-B/142026.05 | 70.31 | 79.55 | 4.56 | |
| COMPOSE-CTBackbone=DINOv2 ViT-B/142026.05 | 69.72 | 79.49 | 6.51 | |
| F-OALBackbone=DINOv2 ViT-B/142026.05 | 67.22 | 77.35 | 0.65 | |
| FeCAMBackbone=DINOv2 ViT-B/142026.05 | 66.95 | 77.7 | 2.28 | |
| RanPACBackbone=DINOv2 ViT-B/142026.05 | 63.99 | 76.29 | 2.28 |