Class-Incremental Learning on CUB-200 Split (10-task)
88FAAJoint-Train.
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Joint-Train.Backbone=ViT-Base, Pre-training=ImageNet-21K2024.10 | 88 | — | |
| VQ-PromptBackbone=ViT-Base, Pre-training=ImageNet-21K, Backbone status=Frozen2024.10 | 86.72 | 90.33 | |
| HiDe-PromptBackbone=ViT-Base, Pre-training=ImageNet-21K, Backbone status=Frozen2024.10 | 86.61 | 87.01 | |
| SLCABackbone=ViT-Base, Pre-training=ImageNet-21K, Backbone status=Not frozen2024.10 | 84.71 | 90.94 | |
| BiCBackbone=ViT-Base, Pre-training=ImageNet-21K, Backbone status=Frozen2024.10 | 81.91 | 89.92 | |
| CODA-PBackbone=ViT-Base, Pre-training=ImageNet-21K, Backbone status=Frozen2024.10 | 70.03 | 74.26 | |
| LwFBackbone=ViT-Base, Pre-training=ImageNet-21K, Backbone status=Frozen2024.10 | 69.75 | 80.45 | |
| DualPromptBackbone=ViT-Base, Pre-training=ImageNet-21K, Backbone status=Frozen2024.10 | 66 | 77.92 | |
| FT++Backbone=ViT-Base, Pre-training=ImageNet-21K, Backbone status=Frozen2024.10 | 37.81 | 63.55 | |
| FTBackbone=ViT-Base, Pre-training=ImageNet-21K, Backbone status=Frozen2024.10 | 11.04 | 31.96 |