现代电子技术2024，Vol.47Issue(15)：73-80,8.DOI:10.16652/j.issn.1004-373x.2024.15.012
基于改进EfficientNetB0模型的葡萄叶部病害识别方法
Grape leaf disease identification method based on improved EfficientNetB0 model
摘要
Abstract
In order to efficiently and accurately identify grape leaf diseases,an LE-EfficientNet model is proposed.On the basis of the EfficientNetB0 model,the large kernal attention(LKA)mechanism is used to replace the squeeze-and-excitation network(SENet)in the MBConv module of the original model.Then,skip connection is used to integrate efficient channel attention(ECA)mechanism behind the last convolutional layer,which makes the network more efficient in extracting local important information of grape leaf diseases in combination with the three attention mechanisms.The SGD(stochastic gradient descent)optimizer of the original model is replaced with Adam optimizer,which improves the generalization ability of the classification model.Training is carried out on the grape leaf disease dataset PlantVillage,and the results show that the accuracy rate of LE-EfficientNet model is improved by 1.58%,its overall accuracy is increased by 1.62%,its recall rate is increased by 1.46%,and its F1-score is increased by 1.53%in comparison with those of the original model.In addition,its parameter quantity is only 10.18 MB,which is decreased by 2.7 MB in comparison with that of the original model.In comparison with the other classical network models,the performance evaluation indexes of the proposed model are improved to varying degrees.To sum up,this study provides a new reference for the identification of grape leaf diseases.关键词
葡萄叶部病害/卷积神经网络/图像分类/大核注意力机制/高效通道注意力机制/EfficientNetB0Key words
grape leaf disease/convolutional neural network/image classification/LKA mechanism/efficient channel attention mechanism/EfficientNetB0分类
信息技术与安全科学引用本文复制引用
胡施威,邓建新,王浩宇,邱林..基于改进EfficientNetB0模型的葡萄叶部病害识别方法[J].现代电子技术,2024,47(15):73-80,8.基金项目
国家自然科学基金面上项目(32270022) （32270022）
