Automatic Classification of Wheat Rust Diseases Using Deep Convolutional Neural Networks

Wheat is the staple food for Indians and it is one of the most common grain crops all over the world. The wheat diseases cause a huge amount of yield losses. The wheat yield losses are due to fungi, bacterial and insect-based diseases. Due to fungal diseases, yield loss has a great effect on wheat grain. Based on the type of fungus, the fungal diseases are categorized into four types namely as rust, leaf spot, spike infection, and virus-based diseases. Hence, these rust diseases affect the whole wheat plant and lead to very heavy yield quality loss. To overcome these quality yield issues, a deep learning-based approach known as a deep convolutional neural network (DCNN) is proposed which can easily classify the wheat rust diseases automatically without human investigation. Additionally, this DCNN training and testing process produces wheat rust diseases determination and high classification results. A total of 1486 wheat plant images have been accessed through the CGIAR dataset and 514 wheat stem rust images have been collected from secondary sources. A total of 2000 wheat plant images have been used for training and testing purposes in DCNN. During training, a stochastic gradient descent method achieves high classification results. Therefore, our proposed approach achieves high classification accuracy of 97.16% for wheat rust diseases.