Realization of a Yolo-V3-Based Application for the Detection of Tick-Borne Cases

Abstract

Today, tick-borne diseases have become widespread and pose a significant threat to public health. There are various types of diseases transmitted from ticks to humans. The main of these diseases; Crimean-Congo hemorrhagic fever, Lyme disease, Mediterranean spotted fever and Tularemia can be listed. As with other types of diseases, early diagnosis is important in tick-borne diseases. Therefore, it is necessary to identify ticks quickly and accurately in order to reduce the possible risks of disease in cases of errors bitten by ticks. In this study, Yolo-v3-based deep learning algorithm, which is a subfield of machine learning, was used primarily to detect ticks. For the training and testing of this algorithm, a new data set was created by downloading 1500 different tick images from the internet. Algorithm was trained and tested using this data set. In order to determine that the success accuracy of the Yolo-v3-based deep learning algorithm is superior and to demonstrate its availability in real life, various performance tests were performed and an estimate was made as to whether there were ticks in an image. As a result of the study, in order to reduce the disease risk of patients bitten by ticks and to intervene in a timely manner, tick detection was made effectively by taking only one tick picture.