Implementasi Metode YOLO V8 untuk Klasifikasi Jenis dan Tingkat Keparahan Jerawat Berbasis Android

Abstract

Acne is an inflammatory skin disorder that is experienced by almost everyone in the world. Clinically, acne is chronic and self-limiting. Acne is generally triggered by internal factors such as hormones, sebum secretion and hereditary genetic factors and external factors triggered by Propionibacterium (P.Acne) bacteria, stress, environment and chemical contradictions in cosmetics and medicines. The spread of acne disorders with a high prevalence rate in adolescence is not fully supported by education related to acne management and the limitations of the community to visit experts to get diagnoses related to acne suffered. Therefore, to help overcome this problem, a system is needed that can make it easier for people to get an initial diagnosis related to identifying the type of acne and the severity that is being suffered along with education related to acne and its treatment. In this study, 4 types of acne that are commonly suffered can be identified, namely comedo, papule, pustule, and nodule as well as three levels of severity that can be classified as mild, moderate, and severe. The system was built in this study using the You Only Look Once (YOLO) algorithm version 8. YOLOv8 is a detection model algorithm that has improved in combining speed and accuracy in recognising objects in images in realtime. This research will use a dataset of 1,455 data used for both identification and classification tasks. The system built in this study can identify the four types of acne very well with an accuracy rate of 93.24% and classify the severity of acne at an accuracy of 96.59%. Both systems can be used in realtime on android devices.