Facial Skin Analysis for Detection of Dark Circles and Acne

Abstract

The aim of this research is to analyze and detect the most commonly found facial skin conditions like acne vulgaris and dark circles. In the skin analysis algorithm, the occurrences of facial acne and dark circles are detected by using existing Computer Vision algorithms such as Otsu’s thresholding algorithm and Deep Learning (DL) techniques. These techniques are further modified to suit the prepared dataset and achieve greater value of evaluation metrics. This article proposes two techniques for the detection of dark circles, which are the difference in Skin Tone Pixel Values and the Thresholding Technique and compare their performance. In the Thresholding Technique, the IoU obtained was 0.737, which provided better visualization of the affected region. Further, acne detection was carried out using two deep learning backbones viz, Inception ResNet 50 and MobileNet. The accuracy obtained for both the methods was 99%.