Shallot Leaf Disease Detection Using Convolutional Neural Network and Support Vector Machine

Abstract

Shallot is one of the important horticultural commodities in Indonesia with high economic value and increasing market demand. However, shallot productivity often decreases due to plant diseases, especially purple blotch and moler disease. Manual disease identification through visual observation still has limitations because it depends on farmers’ experience and may lead to diagnostic errors. This study aims to compare the performance of Convolutional Neural Network and Support Vector Machine methods in detecting shallot leaf diseases based on leaf images and to implement the best-performing model into a Flask-based web application. The dataset used in this study consists of 520 shallot leaf images divided into two classes: purple blotch and moler. The data were collected from primary sources through direct image acquisition in agricultural fields and secondary sources from Kaggle and Roboflow. The research stages include literature study, data collection, preprocessing, image resizing, data annotation, dataset splitting, Convolutional Neural Network model training, Support Vector Machine training, model evaluation, and Flask web implementation. The results show that the Convolutional Neural Network model achieved a training accuracy of 98% and a validation accuracy of 98%, while the Support Vector Machine model achieved an accuracy of 95%. These results indicate that both methods can effectively detect shallot leaf diseases. The Flask web implementation can also assist users in detecting diseases more practically and quickly