Pengembangan Sistem Deteksi Penyakit Daun Pisang Untuk Meningkatkan Kualitas Ekspor Menggunakan Algoritma Yolov5

Abstract

Indonesia is an agrarian country rich in agricultural and plantation resources. One of the main focuses of this research is the Banana Tree Plant. The Banana Tree is one of the abundant tropical plants in Indonesia. Besides its fruit production as a food source, banana leaves also serve as a significant commodity in the agricultural sector, contributing to economic development. To meet the quality standards and criteria set by export destination countries, product quality is crucial for maintaining market access and income levels. However, one of the primary challenges in banana cultivation is pest attacks, such as Erionota Thrax L., which often disturb and can reduce both yield and harvest quality. The implementation of conventional methods requires significant time and resources. To overcome these challenges, a detection system capable of fast and accurate classification is necessary. One deep learning method that can be utilized is You Only Look Once version 5 (YOLOVS) The YOLOVS algorithm has the ability to recognize objects in images and videos that can be classified in real-time. The data used in this research consists of 407 primsary images trained with various hyperparameters, including epoch, batch size, optimizer, and the YOLOS model. The analysis results indicate that the best texting results were achieved with an epoch value of 50, batch size of 8. Stochastic Gradient Descend (SGD) optimizer, and YOLOvSx model, with a mean Average Precision (AP) sular of 0.567. Thus, farmers can take timely preventive measures, enhance product quality and harvest yields, providing robust support for the sustainability and well-being of farmers in the banana industry.