Class-incremental image classification on NICO++ U (test)
98.03Accuracy (T1)FedET
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| FedETBackbone=ResNet-182026.02 | 98.03 | 62.85 | 51.71 | 43.24 | 47.98 | 43.62 | |
| FedAvgBackbone=ResNet-182026.02 | 96.99 | 64.17 | 50.05 | 42.82 | 49.98 | 39.86 | |
| FedProxBackbone=ResNet-182026.02 | 96.99 | 64.22 | 50.05 | 42.82 | 49.98 | 39.83 | |
| FedEWCBackbone=ResNet-182026.02 | 96.09 | 64.22 | 50.43 | 42.69 | 50.8 | 40.09 | |
| FedIL+Backbone=ResNet-182026.02 | 92.66 | 66.58 | 63.14 | 54.76 | 51.32 | 49.98 | |
| OSCAR-ILBackbone=ResNet-182026.02 | 89.96 | 50.79 | 32.54 | 33.34 | 28.9 | 25.96 | |
| OSCAR-RBackbone=ResNet-182026.02 | 89.96 | 50.35 | 31.64 | 33.58 | 28.94 | 26.03 | |
| OSI-FLBackbone=ResNet-18, Retained samples p=52026.02 | 89.96 | 72.67 | 70.38 | 65.52 | 58.68 | 58.88 |