DeepSkin: Mobile-based Skin Quality Assessment with Deep Learning

Loading...
Thumbnail Image

Authors

Mueller, Will

Date

Type

thesis

Language

eng

Keyword

Skin Quality , Dermoscopy , Deep Learning , Machine Learning , Frailty , Dry , Hair

Research Projects

Organizational Units

Journal Issue

Alternative Title

Abstract

We present a novel skin quality assessment system with the goal of improving skin care. This system leverages digital dermoscopy and consists of hardware, software, and a machine learning component to automate the process of skin quality assessment. We first give a brief introduction to the problem followed by and in-depth overview of skin assessment and related works. Following this, the different components of the system are described along with the data collection and algorithms. We performed two studies with 114 participants in total. These studies yielded 2332 skin images that were ranked by experts across a number of skin quality metrics. We plan to release this dataset as part of this work. Using this data we trained several classical and deep learning algorithms to classify Dryness, Frailty, and existence of hair for the images. The best performance for Dryness detection was achieved by VGG16 with an accuracy of 0.8606, while Frailty was best classified by VGG19 with an accuracy of 0.7929. Additionally, Hair was best detected by VGG16 with an accuracy of 0.8756 We believe our proposed and developed system creates a new research direction in the area of digital mobile-based dermoscopy, showing the feasibility of low-cost, mobile, and reliable skin quality assessment.

Description

Citation

Publisher

License

Queen's University's Thesis/Dissertation Non-Exclusive License for Deposit to QSpace and Library and Archives Canada
ProQuest PhD and Master's Theses International Dissemination Agreement
Intellectual Property Guidelines at Queen's University
Copying and Preserving Your Thesis
This publication is made available by the authority of the copyright owner solely for the purpose of private study and research and may not be copied or reproduced except as permitted by the copyright laws without written authority from the copyright owner.
Attribution 3.0 United States

Journal

Volume

Issue

PubMed ID

External DOI

ISSN

EISSN