Low-shot Classification on ImageNet (unseen)
83.6Top-5 AccuracyFEAT
Evaluation results
| Method | Links | |
|---|---|---|
| FEATBackbone=ResNet-50, Shots=20-Shot2018.12 | 83.6 | |
| PMNBackbone=ResNet-50, Shots=20-Shot2018.12 | 82.6 | |
| FEATBackbone=ResNet-50, Shots=10-Shot2018.12 | 81.2 | |
| PMNBackbone=ResNet-50, Shots=10-Shot2018.12 | 80.1 | |
| ProtoNetBackbone=ResNet-50, Shots=20-Shot2018.12 | 80 | |
| ProtoNetBackbone=ResNet-50, Shots=10-Shot2018.12 | 78.1 | |
| FEATBackbone=ResNet-50, Shots=5-Shot2018.12 | 76 | |
| PMNBackbone=ResNet-50, Shots=5-Shot2018.12 | 75.9 | |
| ProtoNetBackbone=ResNet-50, Shots=5-Shot2018.12 | 74.4 | |
| FEATBackbone=ResNet-50, Shots=2-Shot2018.12 | 65.4 | |
| PMNBackbone=ResNet-50, Shots=2-Shot2018.12 | 65.2 | |
| ProtoNetBackbone=ResNet-50, Shots=2-Shot2018.12 | 64 | |
| FEATBackbone=ResNet-50, Shots=1-Shot2018.12 | 53.8 | |
| PMNBackbone=ResNet-50, Shots=1-Shot2018.12 | 53.3 | |
| ProtoNetBackbone=ResNet-50, Shots=1-Shot2018.12 | 49.6 |