Pathogen Detection in Plant Leaves using Deep Neural Networks with Vision Transformer Technique

Abstract

Better yields can only be achieved by the prompt diagnosis and treatment of crop diseases. Due to its remarkable success in machine vision, Deep Neural-Network (DNN) model with vision transformer has been used in this research to detect and diagnose plant illnesses via their leaves. The approach used in this implementation was trained on plant Village dataset (includes data for 14 plant species, 38 disease classifications, and healthy plant leaves). Parameters including batch size, dropout, and the total number of epochs were used to measure the accuracy of the models. Results from the disease identification accuracy has shown the DNN model with vision transformer obtained an accuracy of 99.67% in plant leaf disease classification, which was significantly higher than the accuracies achieved by the other models in the comparative analysis study. This suggests that the DNN model has a promise for use in the real-time detection of illnesses in agricultural systems.