Real-time Identification of Rice Leaf Diseases using Convolutional Neural Networks

Abstract

Rice is one of the most important crops worldwide, serving as a primary food source for millions of people. However, this crop is threatened by diseases such as Rice Blast, Brown Spot, and Bacterial Blight, which manifest in the leaves of the plant. The characteristics of these diseases, captured in digital images, can be utilized in computer vision techniques for their detection and classification. In this study, two Convolutional Neural Networks, YOLO version 8 and Faster R-CNN, were compared to detect and classify the diseases Blast and Brown Spot in a dataset comprising 3636 images with 7915 annotations indicating the location of the disease on the rice leaves. The model trained using YOLO version 8 achieved an accuracy of 92.98% and a recall of 92.45%, while the model trained with Faster R-CNN achieved an accuracy of 91.99% and a recall of 87.78%. YOLO has lower inference times compared to Faster R-CNN due to its more efficient approach and simpler architecture.