Our embedding function refers to the S2M2_R, and its source code is available at: https://github.com/nupurkmr9/S2M2_fewshot.
pleast refer to it for the backbone training.
You can run "save_plk.py" for the embedding extraction.
You can make evaluation through the "Learning2Capture.py".
@inproceedings{
Chi2021Learning,
title={Learning to Capture the Query Distribution for Few-Shot Learning},
author={Chi, Ziqiu, and Wang, Zhe, and Yang, Mengping, and Li, Dongdong, and Du, Wenli},
year={2021},
}
https://arxiv.org/abs/2101.06395
https://github.com/ShuoYang-1998/Few_Shot_Distribution_Calibration