Class Incremental Learning on ImageNet-R Inc5 (test)
72.52Average AccuracyCoRe
Evaluation results
| Method | Links | ||
|---|---|---|---|
| CoReBackbone=ViT-B/16-IN21K2026.03 | 72.52 | 64.4 | |
| AdapterBackbone=ViT-B/16-IN21K2026.03 | 72.35 | 64.33 | |
| FinetuneBackbone=ViT-B/16-IN21K2026.03 | 70.51 | 62.42 | |
| SSFBackbone=ViT-B/16-IN21K2026.03 | 68.94 | 60.6 | |
| PromptBackbone=ViT-B/16-IN21K2026.03 | 68.79 | 60.48 | |
| ViTBackbone=ViT-B/16-IN21K2026.03 | 62.58 | 54.55 | |
| SimpleCILBackbone=ResNet-101, Initial Classes (B-m)=0, Incremental Protocol (Inc-n)=52026.03 | 8.06 | — | |
| G-ACILBackbone=ResNet-101, Initial Classes (B-m)=0, Incremental Protocol (Inc-n)=52026.03 | 7.63 | — | |
| RanPACBackbone=ResNet-101, Initial Classes (B-m)=0, Incremental Protocol (Inc-n)=52026.03 | 7.11 | — | |
| AnaCPBackbone=ResNet-101, Initial Classes (B-m)=0, Incremental Protocol (Inc-n)=52026.03 | 7.01 | — | |
| KLDABackbone=ResNet-101, Initial Classes (B-m)=0, Incremental Protocol (Inc-n)=52026.03 | 6.89 | — | |
| LoRanPACBackbone=ResNet-101, Initial Classes (B-m)=0, Incremental Protocol (Inc-n)=52026.03 | 6.66 | — | |
| SCL-MGSMBackbone=ResNet-101, Initial Classes (B-m)=0, Incremental Protocol (Inc-n)=52026.03 | 6.42 | — |