Unsupervised Video Class-Incremental Learning on Something-to-Something V2 (SSv2)
12.15Avg AccuracyuVCIL-CLU
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| uVCIL-CLUBackbone=ResNet-342026.01 | 12.15 | -4.12 | 0.76 | |
| uVCIL-CLU-RBFBackbone=ResNet-342026.01 | 11.53 | -4.02 | 0.89 | |
| iCaRLBackbone=ResNet-342026.01 | 11.37 | -7.23 | 0.99 | |
| iCaRL+CLBackbone=ResNet-342026.01 | 9.87 | -4.76 | 0.86 | |
| iCaRL+BiCBackbone=ResNet-342026.01 | 9.45 | -5.03 | 1.01 | |
| CLU-RBF (All in one)Backbone=ResNet-342026.01 | 7.23 | — | — | |
| CLU (All in one)Backbone=ResNet-342026.01 | 7.09 | — | — | |
| EWCBackbone=ResNet-342026.01 | 6.6 | -4.55 | 1.03 | |
| MASBackbone=ResNet-342026.01 | 6.52 | -4.46 | 0.88 |