Video Classification on Kinetics (meta-test)
0.73Accuracy (1-shot)TAM
Evaluation results
| Method | Links | ||
|---|---|---|---|
| TAMn-way=5-way, Protocol=Fine-tuned on meta-training set, Backbone=ResNet-50, Episodes=100002019.06 | 0.73 | 0.858 | |
| TRN++n-way=5-way, Protocol=Fine-tuned on meta-training set, Backbone=ResNet-50, Episodes=100002019.06 | 0.684 | 0.82 | |
| CMN++n-way=5-way, Protocol=Fine-tuned on meta-training set, Backbone=ResNet-50, Episodes=100002019.06 | 0.654 | 0.788 | |
| TSN++n-way=5-way, Protocol=Fine-tuned on meta-training set, Backbone=ResNet-50, Episodes=100002019.06 | 0.645 | 0.779 | |
| CMNn-way=5-way, Protocol=Fixed backbone (ImageNet pre-trained), Backbone=ResNet-50, Episodes=100002019.06 | 0.605 | 0.789 | |
| MAMLn-way=5-way, Protocol=Fixed backbone (ImageNet pre-trained), Backbone=ResNet-50, Episodes=100002019.06 | 0.542 | 0.753 | |
| Matching Netn-way=5-way, Protocol=Fixed backbone (ImageNet pre-trained), Backbone=ResNet-50, Episodes=100002019.06 | 0.533 | 0.746 |