Detection of Maize Leaf Diseases Based on MAC Deep Learning

Abstract

Abstract Maize is a globally important food crop, and its yield and quality are vulnerable to various leaf diseases. To address issues such as blurred edges of disease spots and difficulty in small target detection, this study proposes the MAC-YOLO11 model improved on the basis of YOLOv11m. The model introduces the MSPP module to enhance the extraction capability of edge and directional features, and incorporates the spatial position attention module CDSA to strengthen the modeling capability for differences in lesion morphology, texture, and spatial distribution. We designed the APC module to expand the effective receptive field at a low parameter cost through asymmetric convolution branches. The dataset covers 11 categories: northern leaf blight, brown spot, common rust, smut, downy mildew, fall armyworm larval damage, gray leaf spot, maize streak virus disease, adult corn borer damage, corn borer larval damage, and healthy maize leaves. Results show that the MAC deep learning model achieves mAP50 and mAP50:95 of 94.1% and 83.5%, respectively, with overall performance superior to YOLOv11m and other mainstream models. This study provides a technical solution for intelligent identification of maize diseases and holds significant value for disease monitoring and precise control in smart agriculture.

Code and data availability

The paper's authors explicitly state that the source code and implementation details of the proposed MCA-YOLO11 maize leaf disease detection model are publicly available on GitHub, matching an allowed URL. No public dataset deposit is stated for the 17,729-image maize leaf disease dataset.