Classification of Diseases in Paddy using Deep Convolutional Neural Network

Abstract

Abstract Paddy disease detection is decisive in the field of automatic pathogens diagnosis machine. Currently, Deep Con- volutional neural network typically examined the state-of-the art results in image classification. In this work, we proposed a novel DCNN model to identify previously known bacteria leaf blight, brown spot, leaf blast, leaf smut and narrow diseases in prior knowledge. A unique repository of data holds 1260 images of different diseases, 80% of data carried out for training and 20% for testing the samples. To add advantages to our model, we built our model using ADAM optimizer and conducted comparative research over SVM (support vector machine), KNN (K-Nearest neighbor) and ANN (Artificial Neural Network). The dataset given to the novel DCNN model with keras framework and achieved testing accuracy of 0.940 with less training error rate of 0.013. The interpretation outcome demonstrates that high level image classification accuracy with less error rate was achieved by novel DCNN model than traditional methods. Therefore, our model performs best for recognizing 5 paddy diseases and can be possibly implemented in day to day life application.