Free Lunch to Meet the Gap: Intermediate Domain Reconstruction for Cross-Domain Few-Shot Learning

Introduction

This repository contains code for the paper named Free Lunch to Meet the Gap: Intermediate Domain Reconstruction for Cross-Domain Few-Shot Learning.

Quick Started

1. Environment Set Up

Clone this repository and install packages.

Python 3.8
PyTorch 1.7.1 or higher
torchvision 0.8.2 or higher
numpy

2. Download Pretrained Weights

Please download our checkpoints from Google Drive and put it in ./results/.

3. Evaluate on your own dataset.

python meta_test.py --config configs/test.yaml

Train Your Own Data

1. Prepare your own data as the metadata\mini-ImageNet.

2. Start training!

python train.py --config configs/train.yaml
python meta_train.py --config configs/meta_train.yaml

Contacts

If you have any questions about our work, please contact us by email.

Tong Zhang: tongzhang@buaa.edu.cn

Acknowledgments

Our code is build upon Hawkeye and FRN, thanks to all the contributors!

Citation

@article{tong2025free,
      title={Free Lunch to Meet the Gap: Intermediate Domain Reconstruction for Cross-Domain Few-Shot Learning},
      author={Tong Zhang and Yifan Zhao and Liangyu Wang sand Jia Li},
      journal={arXiv preprint arXiv:},
      year={2025}
}