The following datasets are used in our paper:
miniImageNet contains 64 classes for training, 16 classes for validation, and 20 classes for test.
tieredImageNet contains 608 ImageNet classes that are grouped into 34 high-level categories, which furtherare divided into 20/351, 6/97, and 8/160 categories/classes for training, validation, and test.
CIFAR-FS is derived from CIFAR-100 dataset, which is build by randomly splitting 100 classes of the CIFAR-100 dataset into 64, 16,and 20 classes for training, validation, and testing, respectively.
CUB_200_2011 is a fine-grainde dataset consisting of 11,778 images from 200 bird categories, 100/50/50 classes are divided into train/val/test set.
To test in the 5-way K-shot setting:
bash scripts/test/{dataset_name}_5wKs.sh
For example, to test DCAN on the miniImagenet dataset in the 5-way 1-shot setting:
bash scripts/test/miniimagenet_5w1s.sh
To train in the 5-way K-shot setting:
bash scripts/train/{dataset_name}_5wKs.sh
For example, to train DCAN on the CUB dataset in the 5-way 1-shot setting:
bash scripts/train/cub_5w1s.sh