Table 7 L1 distance of predicted attribute score to ground truth labels for 50-way CUB
From: Multimodal few-shot classification without attribute embedding
Backbone | Number of shots | Training data | Test data |
|---|---|---|---|
ResNet-18 | 1-shot | 0.497340 | 0.497224 |
2-shot | 0.498417 | 0.498723 | |
5-shot | 0.498584 | 0.498149 | |
10-shot | 0.499778 | 0.499468 | |
20-shot | 0.500094 | 0.499941 | |
ResNet-101 | 1-shot | 0.000291 | 0.182606 |
2-shot | 0.000276 | 0.187820 | |
5-shot | 0.000275 | 0.192094 | |
10-shot | 0.000298 | 0.190431 | |
20-shot | 0.000293 | 0.194535 |