Class-incremental classification on ImageNet Sub 10 tasks
89.9Average AccuracySeGP-CL (Ours)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SeGP-CL (Ours)Replay=false, Backbone=CLIP-ViT-B/162026.03 | 89.9 | 80.5 | |
| SeGP-CL-onlyCLIPReplay=false, Backbone=CLIP-ViT-B/162026.03 | 88.7 | 78.8 | |
| CLAPPublication=NeurIPS’24, Replay=true, Backbone=CLIP-ViT-B/162026.03 | 87.8 | 79.2 | |
| RAPFPublication=ECCV’24, Replay=false, Backbone=CLIP-ViT-B/162026.03 | 87.5 | 80.2 | |
| MG-CLIPPublication=ICCV’25, Replay=false, Backbone=CLIP-ViT-B/162026.03 | 87.3 | 78.4 | |
| MoE-AdapterPublication=CVPR’24, Replay=false, Backbone=CLIP-ViT-B/162026.03 | 86.4 | 76.7 | |
| DesCLIPPublication=TMM’26, Replay=false, Backbone=CLIP-ViT-B/162026.03 | 86.4 | 77.8 | |
| CLG-CBMPublication=CVPR’25, Replay=false, Backbone=CLIP-ViT-B/162026.03 | 86 | 78.5 | |
| Aper-AdapterPublication=IJCV’24, Replay=false, Backbone=CLIP-ViT-B/162026.03 | 85.8 | 76.4 | |
| DMNSPPublication=ICCV’25, Replay=false, Backbone=CLIP-ViT-B/162026.03 | 85 | 76.1 | |
| PROOFPublication=TPAMI’25, Replay=true, Backbone=CLIP-ViT-B/162026.03 | 84.7 | 72.5 | |
| DualPromptPublication=ECCV’22, Replay=false, Backbone=CLIP-ViT-B/162026.03 | 80.7 | 67.4 | |
| L2P++Publication=CVPR’22, Replay=false, Backbone=CLIP-ViT-B/162026.03 | 80.5 | 67.2 | |
| SLCAPublication=ICCV’23, Replay=false, Backbone=CLIP-ViT-B/162026.03 | 78.6 | 59.9 | |
| Continual-CLIPReplay=false, Backbone=CLIP-ViT-B/162026.03 | 72 | — | |
| CODA PromptPublication=CVPR’23, Replay=false, Backbone=CLIP-ViT-B/162026.03 | 64.1 | 34.8 |