Class-Incremental Learning on Cars-196 B0 Inc10 (test)
83.1Avg AccuracyPAM (RN152)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PAM (RN152)Backbone=ResNet-1522026.03 | 83.1 | 77.23 | |
| PAM (RN101)Backbone=ResNet-1012026.03 | 80.16 | 77.3 | |
| PAM (RN18)Backbone=ResNet-182026.03 | 79.09 | 64.82 | |
| PAM (RN50)Backbone=ResNet-502026.03 | 77.41 | 62.23 | |
| SimpleCILBackbone=ViT-B/16, Pre-trained Model=ImageNet-21K2026.03 | 54.95 | 35.43 | |
| EASEBackbone=ViT-B/16, Pre-trained Model=ImageNet-21K2026.03 | 49.32 | 34.75 | |
| APER-AdapterBackbone=ViT-B/16, Pre-trained Model=ImageNet-21K2026.03 | 47.91 | 35.49 | |
| DualPromptBackbone=ViT-B/16, Pre-trained Model=ImageNet-21K2026.03 | 45.3 | 30.15 | |
| L2PBackbone=ViT-B/16, Pre-trained Model=ImageNet-21K2026.03 | 42.06 | 30.07 | |
| FinetuneBackbone=ViT-B/16, Pre-trained Model=ImageNet-21K2026.03 | 41.9 | 19.47 | |
| CODA-PromptBackbone=ViT-B/16, Pre-trained Model=ImageNet-21K2026.03 | 36.22 | 25.44 |