Cotton Leaf Disease (CLD) Classification Using Multihead External Attention Vision Transformer

Cotton production is seriously threatened by cotton leaf diseases (CLDs), which lower yield and quality. Accurate categorization and timely identification of these diseases are essential for sustainable agriculture and efficient crop management. This research provides an automated method using Multihead External Attention Vision Transformer (MEA-ViT) for CLD classification. This paper proposed an approach that effectively detects and classifies common CLDs like fusarium wilt, bacterial blight, armyworms, curl virus, aphids, and powdery mildew by utilizing visual transformers and external attention techniques. A comprehensive dataset combining two datasets of labeled cotton leaf images was utilised to train the model, ensuring robustness and reliability under real-world conditions. The proposed model in this study attains an accuracy of 98.75%, which shows its effectiveness and superiority over the traditional methods. This automated classification system will help farmers and agronomists adopt environmentally friendly techniques, reduce crop losses, and act promptly.