Rice Doctor: Paddy Disease Detection from Leaf Image Using Machine Learning

Abstract

v ©Daffodil International University ABSTRACT Rice is one of the main foods in the Indian Subcontinent especially in Bangladesh. In Bangladesh, about 135 million peoples staple food is rice. At present, Bangladesh is producing about 25 million tons’ rice to feed her 135 million people. The amount could be more if millions of rice is not wasted every year by disease. As Bangladesh is a developing country, the illiteracy rate has not decreased significantly yet. Most of the farmers are not familiar with the different kinds of disease in rice. So this is really interesting research to identify the disease by infected leaves in agricultural fields. This research paper is a prototype of detection and classification of Paddy disease by infected leaves using machine learning algorithms. We consider three rice diseases named Blast disease, Plant Hopper Disease and Leaf Folder disease. Here we have used a deep learning model based on CNN for classification. We have also used some other transfer learning model for classification. The most important part of this research is to pre-process the image carefully. After processing the data, we trained our model and validated over the dataset. Finally we tested various models but the CNN gives the best result for our dataset, CNN gives an accuracy of about 99.89%.