桂林电子科技大学学报2024，Vol.44Issue(6)：585-591,7.DOI:10.16725/j.1673-808X.202220
一种改进O2U网络的带噪声标签图像分类方法
Image classification method with noisy labels based on improved O2U-Net
摘要
Abstract
In recent years,the research on image classification algorithms with noise labels has attracted wide attention in the aca-demic circle.Overfitting To Underfitting(O2U)network is a learning framework for removing noise labels based on the different loss values of noise label samples in Overfitting and Underfitting states.However,this method faces the risk of incomplete clear-ance of noise label samples.An improved image classification method with noisy labels based on O2U-Net was proposed.By modi-fying part of the loss function of the denoising frame,the network was robust in the denoised data set,and the influence of O2U-Net on removing unclean noise label samples was reduced.Experimental results show that compared with O2U-Net,the proposed robust loss function combined with denoising framework can improve the classification effect on MNIST,CIFAR-10 and CIFAR-100 data sets.The effects of tag noise rate and noise distribution on classification are summarized.The experiment shows that the classifica-tion effect is determined by the noise rate and noise distribution.关键词
图像分类/噪声标签/O2U网络/鲁棒损失函数/去噪算法Key words
image classification/noisy label/O2U-Net/robust loss function/denoising algorithm分类
信息技术与安全科学引用本文复制引用
徐智,杜玉,赵龙阳,孟瑞敏,李沁璘..一种改进O2U网络的带噪声标签图像分类方法[J].桂林电子科技大学学报,2024,44(6):585-591,7.基金项目
国家自然科学基金(61662014) （61662014）
广西自然科学基金(2020GXNSFAA297186) （2020GXNSFAA297186）
广西科技基地和人才专项(AD19110022) （AD19110022）
