Rice Plant Disease Detection and Classification Using Deep Residual Learning

Abstract

In the present scenario, India’s economy highly depends upon the farming output and agricultural productivity. Hence, identifying and detecting diseases in the crops or plants become profoundly important, as it is very natural for plants out there in the fields to get attacked by certain specific bacterial or fungal diseases. If not taken care of at the earliest, this may prove to be a disaster for the product quality and quantity, or one may say productivity overall. For better efficiency at this goal, Machine Learning concepts can definitely be helpful, rather than just by visual sightings and recognition. The following research presents a paradigm for the detection and classification of diseases in rice plants, one of the major crops of the Indian staple diet, using the images of tainted rice plants. Three diseases were mainly focused on namely Bacterial leaf blight, Brown spot, and Leaf smut. The Rice Leaf Disease Dataset, from the UCI Machine Learning Repository, was used. To classify the images into desired disease classes, Residual Neural Network was used which is found to be a speedy, highly efficient technique and gives better results than the plain Convolutional Neural Network and other classifiers such as the Support Vector Machines, by not letting the model to reach saturation level for larger data or deeper networks. We achieved an accuracy of about 95.83% on the dataset.