A Multi facet Deep Neural Network Model for Various Plant Disease Detection

Plant diseases are very common in India, but their effect is severe. If those could be identified beforehand, most of the disease can be checked in their early stage. Thereby their severity could be avoided by taking necessary precautions. In this paper we present a deep learning based neural network model to do disease prediction in various plants. It makes use of a convolutional neural network architecture. For testing the model, an open source data set has been used. A total of 3790 images are collected from more than six plants, which carry more than 11 types of disease. Data Augmentation has been used to strengthen the training dataset. The application of data augmentation has increased the accuracy from 72% to 92%. The advantage of Transfer learning has been exploited in the current work. Google Inception V3 model is used to train the data set. The overall accuracy for this model is coming around 93%, where as for certain disease like Black Measles the proposed model is giving an accuracy up-to 99.35%, which is quite comparable with the other contemporary research works.