Automated Disease Detection in Plant Images using Convolution Neural Network

In this work, authors developed detection system for plant-disease using convolution neural network (CNN) model. The developed system was trained using healthy and unhealthy plants leaves on the dataset of 20,639 images taken in both lab and real-world environment from the open fields. The dataset used in this work includes 3 plant species with 15 dissimilar classes including some healthy leaves of plants. Performance of developed model was compared against most prominent machine learning algorithms such as LDA, KNN, and Gaussian Naïve Bayes classifier etc. On comparing the results, convolution neural network model was found comparatively more competent than other existing machine learning using CNN achieving the peak classification accuracy of 97% and then followed by LDA and KNN with accuracy of 95.7% and 94.4%, respectively. The results reveals that the developed system can effectively be utilized for disease detection and real-time monitoring of plant crops in the open fields.