This paper studies the corn leaf disease detection technology based on the improved YOLOv5 model, aiming to enhance the real-time and accurate identification ability of corn leaf diseases. To address challenges such as multi-scale lesions, small targets, and complex backgrounds, this paper integrates attention mechanisms such as SENet, ECANet, and CBAM into the YOLOv5 architecture, enhancing the model's feature representation ability, especially for the detection of blight disease ，gray leaf spot disease and corn common rust . By training on a dataset containing 4,186 images of corn leaves, the improved model achieved a mAP@0.5 of 91.3% after 200 rounds of training, which was a significant improvement over the latest YOLOv12 model at present. The experimental results show that after integrating the attention mechanism, the model performs well in small-scale disease detection, but there are still certain performance differences for some disease types. The research provides effective technical support for the deep learning detection of agricultural diseases and points out that in the future, directions such as expanding disease classification, enriching datasets, and combining with unmanned aerial vehicle systems for real-time monitoring can be explored to further enhance the efficiency and accuracy of crop health assessment.
Tao Xu (Sat,) studied this question.