Table 5 10-way classification accuracy on AWA2 of proposed model
From: Multimodal few-shot classification without attribute embedding
Backbone | Metric | 1-shot | 2-shot | 5-shot | 10-shot | 20-shot |
|---|---|---|---|---|---|---|
ResNet-18 | Top-1 | 59.67 | 74.79 | 79.12 | 87.19 | 91.47 |
Top-3 | 84.73 | 90.88 | 94.64 | 96.78 | 98.41 | |
Top-5 | 93.13 | 96.67 | 98.08 | 98.91 | 99.60 | |
ResNet-101 | Top-1 | 65.37 | 78.13 | 83.56 | 87.12 | 89.50 |
Top-3 | 87.98 | 92.12 | 95.32 | 97.22 | 98.61 | |
Top-5 | 95.01 | 96.73 | 97.61 | 98.77 | 99.64 |