Article
Deep-Learning-Based Hair Damage Diagnosis Method
Applying Scanning Electron Microscopy Images
Lintong Zhang, Qiaoyue Man and Young Im Cho *
Citation: Zhang, L.; Man, Q.; Cho,
Y.I. Deep-Learning-Based Hair
Damage Diagnosis Method Applying
Scanning Electron Microscopy
Images. Diagnostics 2021, 11, 1831.
https://doi.org/10.3390/
diagnostics11101831
Academic Editor: Keun Ho Ryu and
Nipon Theera-Umpon
Received: 25 August 2021
Accepted: 29 September 2021
Published: 3 October 2021
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2021 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
AI/SC Lab, Computer Engineering of Gachon University, Seongnam-si 461-701, Gyeonggi-do, Korea;
zhanglintong1@naver.com (L.Z.); manqiaoyue@gmail.com (Q.M.)
* Correspondence: yicho@gachon.ac.kr; Tel.: +82-10-3267-4727
Abstract:
In recent years, with the gradual development of medicine and deep learning, many
technologies have been developed. In the field of beauty services or medicine, it is particularly
important to judge the degree of hair damage. Because people in modern society pay more attention
to their own dressing and makeup, changes in the shape of their hair have become more frequent, e.g.,
owing to a perm or dyeing. Thus, the hair is severely damaged through this process. Because hair
is relatively thin, a quick determination of the degree of damage has also become a major problem.
Currently, there are three specific methods for this purpose. In the first method, professionals engaged
in the beauty service industry make a direct judgement with the naked eye. The second way is to
observe the damaged cuticle layers of the hair using a microscope, and then make a judgment. The
third approach is to conduct experimental tests using physical and chemical methods. However,
all of these methods entail certain limitations, inconveniences, and a high complexity and time
consumption. Therefore, our proposed method is to use scanning electron microscope to collect
various hair sample images, combined with deep learning to identify and judge the degree of hair
damage. This method will be used for hair quality diagnosis. Experiment on the data set we made,
compared with the experimental results of other lightweight networks, our method showed the
highest accuracy rate of 94.8%.
Keywords: hair damage; image classification; deep learning; damaged cuticle layers; SEM image
1. Introduction
Hair quality inspection and damage determination in beauty salons rely only on
the judgment of professionals such as beauticians, which is mostly based on the tactile
experience of beauticians and observations with the naked eye. However, in the absence
of excessive professional experience, errors can occur when judging the quality of the
hair. The current methods of judging hair damage are to determine the hair damage by
detecting the moisture content, cystine content, coagulation, and/or relaxation; apply a dye
absorption method, the alkali solubility, copper absorption method, or lithium bromide
method [
1
]; or consider the absorbance and tensile strength. Ref. [
2
] Using these methods,
we need to test the composition of many aspects in a chemical or physical experiment.
Compared with our proposed method, this type of chemical and physical method for
detecting hair is complicated and time-consuming. Therefore, we need a faster and simpler
approach to determine the degree of hair damage. This method will be applied to portable
devices in the future, and everyone can easily judge the degree of damage to one’s hair
anytime and anywhere, and then decide whether to receive a perm, dye, or other beauty
services. To better observe cuticle layers in the hair, we used a scanning electron microscope
(SEM) and observed it at
×
800. Thus, according to the form of cuticle layers (Figure 1),
damaged cuticle layers (the white part) accounts for the proportion of hair. We define a
lifting up of the cuticle edge and an irregular overlay of cuticles without cracks or holes as
weak damage; cracks or holes due to severe lifting up of cuticle layers as damage; exposure
Diagnostics 2021, 11, 1831. https://doi.org/10.3390/diagnostics11101831 https://www.mdpi.com/journal/diagnostics
