摘要
针对糖尿病视网膜病变自动分级问题,文章提出了一种基于卷积神经网络的糖尿病视网膜病变图像分类模型。该模型采用MobileNet和DenseNet 2种结构作为主干网络,并在此基础上引入类别权重函数和注意力机制进行改进。Aptos-2019数据集上的五分类实验结果表明,文章设计的糖尿病视网膜分类模型能够对病变图像进行有效检测,在五分类任务中的准确率达到了0.8310。
Aiming to address the issue of automatic classification of diabetic retinopathy,this paper introduces model for classifying diabetic retinopathy images based on convolutional neural network.The model utilizes two structures,MobileNet and DenseNet,as the backbone networks,and incorporates a category weight function and attention mechanism for enhancement.The experimental results on the Aptos-2019 dataset,which consists of five categories,demonstrate that the diabetic retinopathy classification model proposed in this paper can efficiently identify lesion images,achieving an accuracy of 0.8310 in the five-category task.
作者
周凯
陈清辉
代壮壮
ZHOU Kai;CHEN Qinghui;DAI Zhuangzhuang(School of Information Engineering,Jiangxi University of Science and Technology,Ganzhou 341000,China)
出处
《无线互联科技》
2024年第15期17-20,共4页
Wireless Internet Science and Technology
