Maize leaf disease classification using convolutional neural network

Abstract

Agriculture faces a variety of maize diseases that farmers are unable to identify them. Diseases grow from day to day and many crops will die due to lack of proper treatment and also failing in finding what kind of disease it has been. The most common diseases in maize are Common rust, Northern leaf grey spot, Blight etc. Examining the plant with bare eyes and identifying the disease will result in imprecise detection of diseases. This, in turn will lead to the inappropriate usage of pesticide and causes harmful chronic diseases to human beings. So, automatic and accurate identification of disease is most essential in food security. Society can produce enough food to meet the demand using recent technologies. The application of digital technologies may save time and protect crops from decaying well in advance. Hence, an idea for detecting the disease in the affected maize crops automatically using recent digital technologies will be of more useful to the farmers. Deep learning has recently grabbed the attention of many researchers and helped to develop an automatic and accurate system for image classification. The deep learning techniques and its variants have great potential in the detection of maize disease in modern agriculture. The main focus of this article will be on fine-tuning and evaluation of deep Convolutional Neural Networks (CNN) for image-based maize leaves disease classification. In this work, a Deep CNN has used to detect and classify the diseases in maize leaf and in order to increase the accuracy of detection, AlexNet architecture is used to detect maize leaf disease. In CNN, the accuracy is 87% and in AlexNet architecture, the accuracy of 98.5% is achieved.