Recognizing Acne Vulgaris Severity Levels: An Application of Faster-CNN and YOLO Methods on Medical Images Flasma Veronicha Hendryanna (a), Yan Watequlis Syaifudin (a*), Muhammad Afif Hendrawan (a), Nobuo Funabiki (b), Indrazno Siradjuddin (c)
a) Department of Information Technology, State Polytechnic of Malang
Jalan Sukarno Hatta no. 9, Malang, 65141, Indonesia
*qulis[at]polinema.ac.id
b) Department of Electrical and Communication Engineering, Okayama University
1 Chome-1-1 Tsushimanaka, Kita Ward, Okayama, 700-8530, Japan
c) Department of Electrical Engineering, State Polytechnic of Malang
Jalan Sukarno Hatta no. 9, Malang, 65141, Indonesia
Abstract
Acne Vulgaris is a form of acne that most commonly affects around 85% of adolescents. Most of them need treatment immediately by a dermatologist because it can occur unavoidable scar after severe inflammatory acne. In addition, acne diagnosis by a dermatologist could be difficult and time-consuming, so computer vision technology can be a good approach for the solution using any optical sensors, such as the digital camera or smartphone camera. This study implements two deep learning methods, namely Faster R-CNN and YOLO, to recognize acne objects from images and classify them into some severity levels that can be used by dermatologists for consistently assessing clinical practice trials. The method comparison results show that the Faster R-CNN model achieves better accuracy than YOLO for acne object detection and severity classification. To provide a user interactive system, a web application has been applied to be used by patients or dermatologists. Moreover, the results in Acne Vulgaris severity levels recognition have been tested and confirmed by dermatologist experts.
Keywords: Acne Vulgaris, computer vision, Faster R-CNN, YOLO, deep learning
Topic: Artificial intelligent and Its Applications
