Classifying the Usual Leaf Diseases of Paddy Plants in Bangladesh Using Multilayered CNN Architecture

Abstract

More than 130 million people in Bangladesh depend on rice as their main food. Half of the employment of the rural area and the agricultural GDP of Bangladesh depend on rice production. Nearly more than 10 million farmer families cultivate rice in Bangladesh. Almost 10% of rice cultivation is depreciated by different types of rice plant diseases caused by pests. This is the reason why we worked on detecting rice plant (Oryza sativa) disease by visual observation(images). 3265 images of rice plant disease have been collected for this study which belongs to four classes: hispa, brown spot, leaf blast and the healthy ones. The images of a diseased leaf are collected from rice fields. We have used a pre-trained model for classification and extracting features. By using this model, we got incentive results with the accuracy rate of 93.21%.