Automatic Detection of Rice Diseases using Deep Convolutional Neural Networks with SGD and ADAM

Rice diseases are the major problem in all over the world of agriculture sector. The early detection of this disease will prevent the huge economic loss for the farmer. This paper proposes a deep learning algorithm to classify the disease in the rice plant. Images of healthy and blast disease affected leaves are taken for the proposed system. The features are extracted for the healthy and brown spot, leaf blast, hispa disease of the rice leaf. The total data set is divided for training and testing purposes. These images are processed with the proposed multi-classification method and the leaf is categorized into brown spot, leaf blast, hispa disease and healthy. In this work CNN model is employed using two different optimizers i.e. SGD and ADAM. It is found that SGD performed well as compared to ADAM for our application as the training accuracy is obtained from proposed model with ADAM optimizer is 92.04 % while with SGD produced the training accuracy 96.32 %.The proposed model capable of getting promising accuracy for healthy and brown spot, leaf blast, hispa disease of the rice using multi-classification model.