Official PyTorch Implementation of Self-Filtering.
Paper "Self-Filtering: A Noise-Aware Sample Selection for Label Noise with Confidence Penalization" is accepted to ECCV 2022.
k denotes memory bank size. It can be set as [2,3,4]
T denotes threshold in confidence penalty. For all experiment, we set it as 0.2
For CIFAR-10, warm_up = 10,model = resnet18
For CIFAR-100, warm_up = 30,model = resnet34
python main.py --dataset cifar10 --model resnet18 --batch_size 32 --lr 0.02 --warm_up 10 --num_epochs 100 --noise_mode instance --r 0.2 --k 2 --T 0.2 --gpuid 0
Note that the code refers to DivideMix.
If you have any problem about our code, feel free to contact 1998v7@gmail.com
If you find the code useful, please consider citing our paper:
@inproceedings{wei2022self,
title={Self-Filtering: A Noise-Aware Sample Selection for Label Noise with Confidence Penalization},
author={Wei, Qi and Sun, Haoliang and Lu, Xiankai and Yin, Yilong},
booktitle={European Conference on Computer Vision},
pages={516--532},
year={2022},
organization={Springer}
}