Vision Transformer Based Models for Plant Disease Detection and Diagnosis

Abstract

Plant health is one of the most interesting aspects in the natural cycle, it needs to be conserved to keep the life of the organisms. Several plant diseases could be observed at early stages in the leaf level, where immediate interventions should be taken to prevent the progression of the disease. The use of deep learning has dramatically increased recently, owing to its remarkable performance in multiple applications in different research areas. In this study, we focus on the detection of tomato diseases at the leaf stage using recent deep learning architectures. Several deep learning models are put in comparative experiments to achieve a stable and robust classification performance with high precision that outperforms previous SOTA results. Vision Transformers (ViT) models reported the top classification re-sults, with an accuracy of 96.7%, 98.52%, 99.1% and 99.7%. The research funding will help in the early automatic detection of diseases in the leaf plants, thus providing necessary treatments and maintaining the natural cycle.