Institusion
Universitas Sumatera Utara
Author
Syahputri, Indah (STUDENT ID : 181402033)
(LECTURER ID : 0026028304)
(LECTURER ID : 0107078404)
Subject
Realtime Detection
Datestamp
2022-12-14 04:38:37
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%.