The code repository for "Hierarchy SeparateEMD For Few-Shot Learning" in PyTorch. If you use any content of this repo for your work, please cite the following bib entry:

@inproceedings{separate2022fewshot,
  author    = {Yaqiang Sun and
               Jie Hao and
               Zhuojun Zou and
               Lin Shu and
               Shengjie Hu},
  title     = {Hierarchy SeparateEMD For Few-Shot Learning},
  booktitle="Methods and Applications for Modeling and Simulation of Complex Systems",
  year="2022",
  publisher="Springer Nature Singapore",
  address="Singapore",
  pages="548--560",
  isbn="978-981-19-9198-1"
}

We propose a novel model-based approach to adapt the few shot classifacation task. We denote our method as Hierarchy SeparateEMD.

Experimental results on few-shot learning datasets with ResNet-12 backbone. We report average results with 10,000 randomly sampled few-shot learning episodes for stablized evaluation.

MiniImageNet Dataset

The following packages are required to run the scripts:

• PyTorch-1.4 and torchvision

• Package tensorboardX

• Dataset: please download the dataset and put images into the folder data/[name of the dataset, miniimagenet or cub]/images

The MiniImageNet dataset is a subset of the ImageNet that includes a total number of 100 classes and 600 examples per class. We follow the previous setup, and use 64 classes as SEEN categories, 16 and 20 as two sets of UNSEEN categories for model validation and evaluation, respectively.

To reproduce our experiments, please use train_fsl.py. There are four parts in the code.

• model: It contains the main files of the code, including the few-shot learning trainer, the dataloader, the network architectures, and baseline and comparison models.
• data: Images and splits for the data sets.
• saves: The pre-trained weights of different networks.
• checkpoints: To save the trained models.

Please use train.py and follow the instructions below. The file will automatically evaluate the model on the meta-test set with 10,000 tasks after given epochs.

For example, to train the 1-shot/5-shot 5-way SeparateEMD model with ResNet-12 backbone on MiniImageNet:

$ python train.py  --max_epoch 60 --model_class SeparateEMD  --backbone_class Res12 --dataset MiniImageNet --way 5 --eval_way 5 --shot 1 --eval_shot 1 --query 15 --eval_query 15 --balance 0.01 --temperature 64 --temperature2 64 --lr 0.0002 --lr_mul 10 --lr_scheduler step --step_size 40 --gamma 0.5 --gpu 1 --init_weights ./saves/initialization/miniimagenet/Res12-pre.pth --eval_interval 1 --use_euclidean
$ python train.py  --max_epoch 60 --model_class SeparateEMD  --backbone_class Res12 --dataset MiniImageNet --way 5 --eval_way 5 --shot 5 --eval_shot 5 --query 15 --eval_query 15 --balance 0.1 --temperature 64 --temperature2 32 --lr 0.0002 --lr_mul 10 --lr_scheduler step --step_size 40 --gamma 0.5 --gpu 0 --init_weights ./saves/initialization/miniimagenet/Res12-pre.pth --eval_interval 1 --use_euclidean

We thank the following repos providing helpful components/functions in our work.

• ProtoNet

• MatchingNet

• PFA

• Transformer

• MetaOptNet

• FEAT

• DeepEMD