Cross-Domain Class-Incremental Learning on PACS 3 tasks
76.21Accuracy (SD)FOSTER w/ CORF
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| FOSTER w/ CORFBackbone=ResNet-342026.05 | 76.21 | 52.37 | 60.97 | |
| FOSTERBackbone=ResNet-342026.05 | 72.3 | 45.49 | 53.12 | |
| iCaRL w/ CORFBackbone=ResNet-342026.05 | 71.77 | 50.54 | 57.89 | |
| DER w/ CORFBackbone=ResNet-342026.05 | 71.37 | 49.98 | 57.77 | |
| MEMO w/ CORFBackbone=ResNet-342026.05 | 69.99 | 48.64 | 55.87 | |
| Replay w/ CORFBackbone=ResNet-342026.05 | 68.9 | 49.7 | 56.74 | |
| DERBackbone=ResNet-342026.05 | 68.65 | 46.11 | 53.74 | |
| iCaRLBackbone=ResNet-342026.05 | 68.14 | 44.8 | 52.59 | |
| MEMOBackbone=ResNet-342026.05 | 64.65 | 42.61 | 49.64 | |
| ReplayBackbone=ResNet-342026.05 | 64.21 | 40.44 | 48.15 | |
| TagFex w/ CORFBackbone=ResNet-342026.05 | 61.15 | 37.44 | 43.37 | |
| DS-AL w/ CORFBackbone=ResNet-342026.05 | 60.25 | 40.28 | 44.38 | |
| DS-ALBackbone=ResNet-342026.05 | 57.24 | 35.38 | 41.27 | |
| TagFexBackbone=ResNet-342026.05 | 56.83 | 33.21 | 39.44 | |
| FineTune w/ CORFBackbone=ResNet-342026.05 | 52.95 | 43.9 | 47.46 | |
| FineTuneBackbone=ResNet-342026.05 | 52.26 | 37.08 | 42.04 | |
| TRIPSBackbone=ResNet-342026.05 | 49.47 | 21.51 | 28.5 | |
| MSL+MOVBackbone=ResNet-342026.05 | 45.89 | 20.82 | 26.97 |