Leaf Disease Detection Model in Gayo Coffee Plantations Using Deep Learning

Abstract

Coffee is one of the most important tropical plantation commodities, significantly supporting the economy of the Gayo Highlands. Attacks of various diseases can significantly reduce the productivity and quality of Gayo coffee. This study developed a leaf disease detection model in coffee plants using the Convolutional Neural Network (CNN) method. The model developed in this study used two datasets. The first dataset, the Gayo Coffee Leaf Disease (PDKG), comprises 900 images of healthy and diseased leaves collected from Gayo coffee plantations. The acquired images in the PDKG dataset were then preprocessed to improve their image quality. The results of model training and testing on the PDKG dataset showed an accuracy of 0.91. On the public Coffee Leaf Diseases (CLD) dataset, the model achieved an accuracy of 0.95, representing a 7.1% increase compared to previous studies. The resulting model can help local coffee farmers in the Gayo Highlands detect leaf diseases early and manage plant health more efficiently and accurately.