摘要
该文旨在解决传统方法在眼疾病识别中分类准确率低的问题,提出了一种改进的眼疾病识别算法,基于Inception-ResNet v2架构,并引入SENet注意力机制、Ghost模块和空洞空间金字塔池化等技术。通过学习通道相关性和加强对重要特征的关注,显著提高了眼疾病分类的准确率,有效区分常见四种眼疾病数据集。为了进一步提高模型的泛化能力,还引入数据增强技术以减少过拟合。相比Efficient-Net、ResNet和Inception-ResNet等经典深度学习模型,该算法表现更优,为眼疾病早期诊断提供了更准确、高效的方法。
This article aims to solve the problem of low classification accuracy in traditional methods for eye disease recognition.It proposes an improved eye disease recognition algorithm based on the Inception-ResNet v2 architecture,and introduces technologies such as SENet attention mechanism,Ghost module,and atrous spatial pyramid pooling.By learning channel correlation and strengthening attention to important features,the accuracy of eye disease classification has been significantly improved,effectively distinguishing the four common eye disease datasets.At the same time,in order to further improve the generalization ability of the model,this article also introduces data augmentation technology to reduce overfitting.Compared to classic deep learning models such as Efficient-Net,ResNet and Inception-ResNet,this algorithm performs better and provides a more accurate and efficient method for early diagnosis of eye diseases.
作者
陆阳
任世卿
LU Yang;REN Shiqing(Shenyang Ligong University,Shenyang 110158,China)
出处
《电子设计工程》
2024年第20期68-71,共4页
Electronic Design Engineering
