DURIAN DISEASE CLASSIFICATION USING TRANSFER LEARNING FOR DISEASE MANAGEMENT SYSTEM

Abstract

Durian fruit is one of the popular fruits in ASEAN countries, even in European countries, making it a high-potential contributor to economic growth in the agricultural sector. However, durian leaf diseases pose significant challenges in most ASEAN countries, such as Malaysia, Indonesia, the Philippines, and Thailand. Traditionally, the identification of leaf diseases relied on manual visual inspection, a labor-intensive and tedious process. To address this issue, we propose a novel approach for durian leaf disease detection and recognition using vision transformers. We employed well-established deep learning models, VGG-19 and ResNet-9, with carefully tuned hyperparameters including epochs, optimizer, and maximum learning rate. Our results indicate that ResNet-9 achieved an impressive accuracy rate of 99.1% when using the Adam optimizer with a maximum learning rate of 0.001. This breakthrough in automated disease recognition promises to significantly reduce labor costs and time for smallholder farmers, enhancing the sustainability of durian cultivation.