Transforming Agricultural Diagnosis: Vision Transformer-Based Classification of Banana Leaf Diseases for Precision Farming
Raj Gaurang Tiwari
Abstract
Accurate and rapid diagnosis of plant diseases is crucial in precision agriculture to maintain healthy crops and higher yields. To this end, the viability of Vision Transformer (ViT) models, namely ViT-b16 and ViT-l16, for the classification of banana leaf diseases was investigated in this research. A large and diverse dataset of banana leaf images was used to train and test the models. The experimental results of the ViT models were compared with the best-performing deep learning architectures such as Graph Convolutional Networks (GCN), EfficientNet-B3, MobileNet-V2 and ResNet-cut. The ViT-b16 outperforms all other models with an impressive 96.88% accuracy. The research results show that the Vision Transformer architecture is excellent for disease classification on banana leaves. An exciting opportunity for further research has opened thanks to this performance, which can be widely used in precision agriculture. The study lays the foundation for better crop management systems that can detect diseases early and intervene to maximize agricultural production.