Deteksi Penyakit pada Daun Tanaman Tomat dengan Metode You Only Look Once Versi 5 (YOLOV5)

Abstract

Tomato is one of the horticultural crops with high nutritional value. The high nutritional value of tomatoes encourages market demand for tomatoes increase and the production of tomatoes increase too. One of the causes the production of tomatoes decrease is disease attacks on tomato plants, especially on the leaves which cause reduced yields in a period and lead to material losses. To classifying the type of tomato leaf diseases, manual examination was carried out which had weaknesses in terms of limited human senses, relatively long time and insufficient knowledge. Consequently, a method that can aid in classification the tomato leaf disease through images in real time is required. There are two types of diseases classified in this study, namely early blight and late blight and one healthy leaf. In this study, the You Only Look Once (YOLO) version 5 method was used. Yolov5 is a real-time object detection algorithm that excels in both accuracy and speed of inference time. In this study, the total dataset used is 780 images, which is 624 datasets are used for training and 156 datasets are used for testing. This system can classify diseases on tomato leaves in real time, according to the tests. The method used produces a very good accuracy even though it is not perfect at 97.4%.