Learning Task-aware Local Representations for Few-shot Learning, IJCAI 2020
- Python 3
- PyTorch 1.4.0
Please refer to DN4.
DataSet is miniImagenet, CUB, StanfordCar or StanfordDog.
- Train:
python -u trainer.py -c ./config/${DataSet}_Conv64F_5way_1shot.json -d 0
python -u trainer.py -c ./config/${DataSet}_Conv64F_5way_5shot.json -d 0
- Test:
python -u test.py -r ./results/${DataSet}_Conv64F_5way_1shot -d 0
python -u test.py -r ./results/${DataSet}_Conv64F_5way_5shot -d 0
Sorry about the mistakes in the Eq.(4) and the Eq.(7),
the Eq.(4) is a step function, and Eq.(7) is the approximation of the Eq.(4) with the adaptive threshold,
both of them repeatedly introduce the process of the Eq.(5),
the paper after correction is here.
If you use this code for your research, please cite our paper.
@inproceedings{ijcai2020-100,
title = {Learning Task-aware Local Representations for Few-shot Learning},
author = {Dong, Chuanqi and Li, Wenbin and Huo, Jing and Gu, Zheng and Gao, Yang},
booktitle = {Proceedings of the Twenty-Ninth International Joint Conference on
Artificial Intelligence, {IJCAI-20}},
publisher = {International Joint Conferences on Artificial Intelligence Organization},
editor = {Christian Bessiere}
pages = {716--722},
year = {2020},
month = {7},
note = {Main track}
doi = {10.24963/ijcai.2020/100},
url = {https://doi.org/10.24963/ijcai.2020/100},
}
Our code is based on DN4.