Potato Leaf Disease Classification Using Vision Transformer Networks

Potato cultivation is vital for global food security due to the crop's nutritional benefits, but its production is vulnerable to diseases like Alternaria solani (early blight) and Phytophthora infestans (late blight) which can severely impact yield. Traditional disease detection methods are often laborintensive, expensive, and limited in real-time application, creating a need for innovative solutions. This study addresses these challenges by exploring advanced Vision Transformer (ViT) models for the classification of potato leaf health. A comparative analysis was conducted on several ViT variants, including SimpleViT, LeViT, CVT, T2TViT, CrossFormer, XCiT, RegionViT, and NaViT, using a dataset of potato leaf images from Kaggle. The XCiT model emerged as the most effective, achieving a precision of 95.69 %, recall of 95.00 %, and an F 1 -score of 95.64 %. These results surpass those reported in previous studies, showcasing the XCiT model's effectiveness in accurately classifying potato leaf diseases. The findings highlight the potential of Vision Transformer models to enhance potato leaf disease detection and management. Future work should focus on further optimizing model efficiency and exploring hybrid approaches to improve classification accuracy even further. This research contributes to the ongoing efforts to advance plant disease management