A Comparative Analysis of Corn Leaf Disease Classification Using Advanced CNN Architectures

The agricultural sector is critical to producing high-quality food while also contributing considerably to the growth of economies and populations. Plant diseases, on the other hand, can result in significant crop losses and even species extinction. Early and accurate diagnosis of plant diseases using automated methods can improve food production quality while minimizing economic losses. Deep learning has recently enhanced the accuracy of picture classification and object detection systems. This study focuses on four types of maize leaves that have been damaged by specific diseases: blight, common rust, grey leaf spot, and healthy leaves. A total of 1306 samples were obtained for blight, 574 for common rust, 1146 for grey leaf spot, and 1164 for healthy leaves. Five advanced deep neural networks (DNNs) are used to classify these corn leaf diseases: CNN, AlexNet, SqueezeNet, Xception, and DarkNet. These models' performance is measured using metrics including validation accuracy, test accuracy, precision, recall, F1 score, and area under the curve (AUC). CNN had a classification accuracy of 92.6%, AlexNet had 94.3%, SqueezeNet had 95.7%, Xception had 97.3%, and DarkNet53 had 97.4%. The findings reveal that the deep learning models utilized in this study outperform typical machine learning classifiers at detecting corn leaf illnesses.