Abstract:
Guava is one of the most economically important fruits in tropical and subtropical
regions like Bangladesh, India, Pakistan and many more. However, its production is
significantly impacted by various leaf diseases. In recent years, computer vision
techniques have emerged as effective tools for automating disease detection in plants.
In this study, we propose a novel approach for guava leaf disease classification using
Vision Transformer (ViT), a state-of-the-art deep learning architecture known for its
ability to capture long-range dependencies in images. Our dataset comprises images of
guava leaves categorized into five classes: Healthy, Canker, Dot, Rust, and
Mummification, totaling approximately 4749 images. We first preprocessed the images
and then trained a Vision Transformer model on the dataset. The experimental results
demonstrate the effectiveness of our approach, with a training accuracy of 97.42%,
validation accuracy of 95.60%, and testing accuracy of 97.54%. Furthermore, we
conducted comparative experiments with traditional transfer learning models, including
ResNet50 model (Training accuracy – 98.76%, Testing accuracy – 95.00%, validation
accuracy – 96.10%), VGG19 model (Training accuracy - 90.39%, Testing accuracy -
89%, validation accuracy - 88.77%), EfficientNetB5 model (Training accuracy -
96.49%, Testing accuracy - 94.97%, validation accuracy - 93.33%). While these models
have shown promising performance in various computer vision tasks, our results
indicate that the Vision Transformer outperforms them in terms of both accuracy and
computational efficiency for guava leaf disease classification. The success of our
proposed approach underscores the potential of Vision Transformers in agricultural
applications, particularly in the early detection and management of plant diseases. Our
findings contribute to the ongoing efforts to leverage artificial intelligence for
sustainable agriculture, offering a powerful tool for farmers and researchers to combat
guava leaf diseases and ensure global food security.
