智慧农业导刊2025，Vol.5Issue(14)：33-36,4.DOI:10.20028/j.zhnydk.2025.14.009
基于改进Swin-Transformer的果树病叶分类模型
摘要
Abstract
In recent years,climate change and changes in agricultural activities have increased the frequency and severity of plant diseases,having a major impact on food production and quality safety.Therefore,to ensure food security,timely and accurate detection and diagnosis of plant diseases are crucial.This paper designs a tree disease leaf classification model based on the improved Swin-Transformer,which optimizes features by integrating dual-path attention mechanisms.At the feature processing level,a multi-level processing structure including layer standardization,adaptive pooling,and fully connected classifiers is designed.This composite architecture maintains the advantages of Transformer's global modeling and significantly improves the efficiency of capturing fine-grained pathological features through an attention-guided feature enhancement mechanism.The proposed model achieves greater accuracy than previous convolution and visual transformer-based models.关键词
深度学习/注意力机制/卷积神经网络/植物病害识别/智慧农业Key words
deep learning/attention mechanism/convolutional neural network/plant disease recognition/smart agriculture分类
信息技术与安全科学引用本文复制引用
江华晋,彭东海,余焕杰..基于改进Swin-Transformer的果树病叶分类模型[J].智慧农业导刊,2025,5(14):33-36,4.基金项目
广东省省级大学生创新创业训练计划项目(S202410576017X) （S202410576017X）
