False Smut Disease Detection in Paddy Using Convolutional Neural Network

Abstract

Rice false smut (RFS) is the most severe grain disease affecting rice agriculture worldwide. Because of the various mycotoxins produced by the causal pathogen, Villosiclava virens, epidemics result in yield loss and poor grain quality (anamorph: Ustilaginoidea virens). As a result, the farmers’ main concern is disease management measures that are effective, simple, and practical. Because of this, we look at the image of the RFS to understand and predict this severe grain disease. This research proposes a model based on the Convolutional Neural Network (CNN), widely used for image classification and identification due to its high accuracy. First, we acquire data from actual rice farming fields with high-resolution RFS images. Then, we train and test our model’s performance using actual images to compare and validate it. As a result, our model provides 90.90% accurate results for detecting the RFS in actual photos. Finally, we evaluate and record all of the data for subsequent studies.