Maize Leaf Disease Detection Using Vision Transformers (ViTs) and CNN-Based Classifiers: Comparative Analysis

Maize diseases can cause major economic losses and jeopardize food security. Maize diseases such as Gray Leaf Spot, Blast, and Common Rust have been identified globally. These diseases can be promptly managed and controlled with the help of early detection and correct identification, minimizing losses. Machine learning models, such as convolutional neural networks (CNNs) and vision transformers, have been increasingly used in maize disease detection. These models have shown promising results, with high accuracy rates in identifying various types of maize diseases. Recently, vision transformers have emerged as a potential alternative to CNNs in image classification tasks, including crop disease detection. In this study, four types of vision transformer models (ViT-B/16, ViT-B/32, ViT-L/16, and ViT-L/32) were used in addition to four types of CNN models (DenseNet121, VGG16, ResNet152V2, and InceptionV3) for maize leaf disease classification. The results showed that ViT-B/16 achieved the highest accuracy of 94.51% using the SGD optimizer, outperforming the other models.