Plant disease detection using convolutional neural network

Abstract

Rice is a staple crop of Bangladesh and many metric tons of it are being destroyed every year due to diseases. If the diseases can be efficiently and accurately classified ed and recognized at early stage, the farmers can get the required help resulting in better rice crop yields. Thus, in an attempt to better increase the rice crop, yield our proposal is to make a website prototype system by using different machine learning algorithms to analyze and recognize different rice crop diseases. By utilizing CNN and its variations for the detection of rice plant diseases, we aim to guide individuals and assist farmers in identifying the infected plants early. By doing so, automated systems can be made to find out the infected crops and suggest diagnosis based on the problems. The photographs of rice plant leaves are taken for brown spot, Hispa and leaf blast diseases. We have used Convolution Neural Network (CNN) which comprises of different layers which are used for prediction. In addition, we have implemented other 4 CNN structures such as GoogleNet, RestNet-152 and VGG19 which is 19-layer deep structure. On the other hand, the features from the infected area are extracted using Histogram Oriented Gradient (HOG) features and for distinguishing between their category these features were given to the Support Vector Machine (SVM). To sum up, by experimentation we will be able to conclude which structure or algorithm has the most success rate. As a result, by this approach the information will be provide at the initial stage so that one can take necessary steps at the beginning to prevent the rice plant diseases and minimize the loss of production.