5-way few-shot classification on tieredImageNet (meta-test)
80.62AccuracyMetaOptNet-SVM(+EHG)
Evaluation Results
| Method | Links | |
|---|---|---|
| MetaOptNet-SVM(+EHG)shot count=5-shot, backbone=ResNet-12, feature dimension=160002026.02 | 80.62 | |
| MetaOptNet-RRshot count=5-shot, backbone=ResNet-12, feature dimension=160002026.02 | 79.21 | |
| MetaOptNet-SVMshot count=5-shot, backbone=ResNet-12, feature dimension=160002026.02 | 79.11 | |
| ProtoNet(+EHG)shot count=5-shot, backbone=ResNet-12, feature dimension=160002026.02 | 78.93 | |
| MetaOptNet-RR(+EHG)shot count=5-shot, backbone=ResNet-12, feature dimension=160002026.02 | 78.93 | |
| ProtoNetshot count=5-shot, backbone=ResNet-12, feature dimension=160002026.02 | 78.24 | |
| MetaOptNet-SVM(+EHG)shot count=5-shot, backbone=4-layer conv, feature dimension=16002026.02 | 71.85 | |
| ProtoNet(+EHG)shot count=5-shot, backbone=4-layer conv, feature dimension=16002026.02 | 71.4 | |
| MetaOptNet-RR(+EHG)shot count=5-shot, backbone=4-layer conv, feature dimension=16002026.02 | 70.95 | |
| MetaOptNet-SVMshot count=5-shot, backbone=4-layer conv, feature dimension=16002026.02 | 70.65 | |
| ProtoNetshot count=5-shot, backbone=4-layer conv, feature dimension=16002026.02 | 69.9 | |
| MetaOptNet-RRshot count=5-shot, backbone=4-layer conv, feature dimension=16002026.02 | 69.84 | |
| MetaOptNet-SVM(+EHG)shot count=1-shot, backbone=ResNet-12, feature dimension=160002026.02 | 64.1 | |
| ProtoNet(+EHG)shot count=1-shot, backbone=ResNet-12, feature dimension=160002026.02 | 63.38 | |
| MetaOptNet-RR(+EHG)shot count=1-shot, backbone=ResNet-12, feature dimension=160002026.02 | 63.38 | |
| MetaOptNet-RRshot count=1-shot, backbone=ResNet-12, feature dimension=160002026.02 | 58.7 | |
| MetaOptNet-SVMshot count=1-shot, backbone=ResNet-12, feature dimension=160002026.02 | 58.63 | |
| ProtoNetshot count=1-shot, backbone=ResNet-12, feature dimension=160002026.02 | 57.99 | |
| MetaOptNet-RR(+EHG)shot count=1-shot, backbone=4-layer conv, feature dimension=16002026.02 | 54.84 | |
| MetaOptNet-SVM(+EHG)shot count=1-shot, backbone=4-layer conv, feature dimension=16002026.02 | 54.11 | |
| ProtoNet(+EHG)shot count=1-shot, backbone=4-layer conv, feature dimension=16002026.02 | 53.21 | |
| MetaOptNet-RRshot count=1-shot, backbone=4-layer conv, feature dimension=16002026.02 | 51.74 | |
| MetaOptNet-SVMshot count=1-shot, backbone=4-layer conv, feature dimension=16002026.02 | 50.92 | |
| ProtoNetshot count=1-shot, backbone=4-layer conv, feature dimension=16002026.02 | 50.36 |