Application of Deep Learning Techniques for Predicting Leaf Lesions in KDML105 Rice Variety
Abstract
This study applies deep learning techniques to predict rice leaf diseases in the KDML105 (Jasmine Rice) variety, focusing on Narrow Brown Spot Disease and blast diseases, which affect the quality of Thai rice. The research developed a predictive model using the YOLO (You Only Look Once) technique combined with Transfer Learning, trained on images of rice leaves collected from paddy fields in Buriram province. The experimental results showed that the YOLOv8 model achieved the highest performance with a mean average precision (mAP) of 0.95 The majority of current research, however, either requires cutting rice leaves for photography or employs close-up shots of the leaves, which is unsuitable for farmers’ everyday use. In order to solve the usability concerns of earlier research, this study attempts to create a long-range rice leaf disease detection system that can capture several leaves in a single wide-angle photo. It also supports sustainable precision agriculture, reduces dependency on chemical usage, and enhances the competitiveness of Thai rice in both domestic and international markets.
Authors: Sittisak Promtara, Watanyoo Suksa-ngiam, Pornchai Mongkolnam
Published in: International Conference on Engineering, Applied Sciences, and Technology (ICEAST) (2025)