A Deep Learning-Based Approach for Zea Mays Leaf Disease Classification

Abstract

Plant diseases are the leading cause of lost agricultural productivity. Most farmers have difficulties in managing and identifying plant infections. Thus, early diagnosis of these diseases would be beneficial for farmers in avoiding future losses. This study provides a comprehensive comparison of different deep-learning algorithms for categorizing maize leaf diseases using image data. Select common type of leaf diseases for classification such as 'Blight', 'Common_Rust', 'Gray_Leaf_Spot', 'Healthy', and 'Phaeosphaeria Leaf Spot'. Five important algorithms, including MobileNetV2, DenseNet, VGG16, VGG19, and InceptionV3, were evaluated based on their performance criteria, like accuracy, precision, recall, and F1-score. Among these algorithms, MobileNetV2 displayed the best accuracy of 93%, beating other models. Precision, recall, and F1-score were also calculated for each illness class, indicating the strengths and shortcomings of each method in disease categorization. The findings emphasize the effectiveness of deep learning algorithms in reliably diagnosing maize leaf diseases, with MobileNetV2 emerging as the top-performing method in the present investigation. These results give useful insights for academics and practitioners wanting to employ deep learning techniques for agricultural disease management and tracking.