Enhancing Plant Disease Classification Accuracy Using a Hybrid CNN–Vision Transformer Model

Abstract

Plant diseases are one of the many factors which reduce agricultural productivity and global food security. Accurate and early diagnosis of plant diseases helps to reduce significant losses to crops and aid in sustainable agriculture. In recent years, deep learning methods for plant disease diagnosis have been of great interest in the field of agriculture. This study proposes a hybrid model, Hybrid plant disease classification using deep learning models, which uses a Convolutional Neural Network (CNN) to obtain local features of the affected plant leaves and a pre-trained Vision Transformer (ViT) to get global features. In this work, the proposed hybrid model is validated using 15 plant disease classes of which different data augmentation techniques such as changing the light exposure and object orientation are employed. Experimental results revealed the convergence stability of the model, the strong generalization ability, and the better accuracy as compared with each of the models used independently.