![Image-based virtual try-on network [42]](https://www.researchgate.net/profile/Hyunwoo-Hwangbo-2/publication/344765109/figure/fig1/AS:948614977318913@1603178843747/Image-based-virtual-try-on-network-42_Q320.jpg)
Content uploaded by Hyunwoo Hwangbo
Author content
All content in this area was uploaded by Hyunwoo Hwangbo on Oct 27, 2020
Content may be subject to copyright.
Available via license: CC BY-NC-ND 4.0
Content may be subject to copyright.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3023040, IEEE Access
1
Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000.
Digital Object Identifier 10.1109/ACCESS.2017.Doi Number
Effects of 3D Virtual “Try-on” on Online Sales
and Customers’ Purchasing Experiences
Hyunwoo Hwangbo1, Eun Hie Kim2, So-Hyun Lee3 and Young Jae Jang4, Member, IEEE
1Department of Global IT Business, Hannam University, Daejoen, 34430 Korea
2Big Data Analytics Team, Kolon Benit co. Ltd., Gwacheon-si, Gyeonggi-do, 18337 Korea
3School of Management, Xi’an Jiaotong University, Xi’an, Shaanxi, China
4Department of Industrial & Systems Engineering, Korea Advanced Institute of Science and Technology (KAIST), Daejeon, 34141 Korea
Corresponding author: Young Jae Jang (e-mail: yjang@kaist.ac.kr).
ABSTRACT The advancement of the Internet and technology has made it possible to purchase and use
different types of products and services online instead of offline. In particular, as the scale of online shopping
malls is rapidly increasing, new functions have been tried and introduced to compensate for the limitation of
not being able to directly wear clothes in an online mall. Among them, 3D virtual try-on is an innovative
service and its technology is being advanced with continuous interest. Technological advances and interest
in 3D virtual try-on have led to a variety of related studies. Most previous studies on virtual try-on have been
conducted on the virtual fitting technology from the perspective of making clothes, or the effects and customer
behavior from the customer perspective. However, there has been no research based on actual data of
customers using 3D virtual try-on to show how virtual try-on affects sales. Therefore, this study understands
the fundamental meaning of virtual ‘try-on’ as a customer experience and examines the effects of 3D virtual
try-on on online sales. We create a 3D body model complemented by adding more diverse body shapes and
sizes, and investigate the effects of virtual try-on on online sales through actual data. In addition, qualitative
data including interviews are used to complement and interpret the results. The results show that virtual try-
on affects the sales results: the average sales per customer increased by 14,000 won (13USD). The most
important finding is that the return rate decreased by 27% by filtering out incorrect sizes and fits. Virtual try-
on may very well replace physical fitting rooms. This study presents an advanced technology of 3D virtual
try-on and shows that virtual try-on is an effective tool to boost sales and decrease customer’s returns in a
case study of women’s casual L brand.
INDEX TERMS virtual try-on (VTO), 3D clothes digitization, customized body, customer experience,
online sales
I. INTRODUCTION
During the past decade, interest has grown in three-
dimensional (3D) virtual ‘try-on,’ and has been recognized for
its value as an effective aspect of customer experience. Its
original intent was to help customers check out size and fit
virtually, but it has become an enjoyable customer experience.
Since virtual try-on can help customers filter out incorrect
sizes and fits to solve the thorny problem of customers’
inability to wear a garment online, it is very useful in online
malls. In addition, a mix-and-match service can provide
hedonic value, and 3D virtual try-on has become a significant
feature of online malls. The development of 3D virtual try-on
has led to hopes that it helps boost online sales, which has
emerged as a compelling retail channel for apparel sales.
According to Smallbizgenius (2019), online retail sales grew
23.3% over 2017 [1]. The online retail sales are expected to
account for 13.7% of global retail sales in 2019. Also, the top
online purchasing categories in 2018 were fashion at 61% [2].
To increase online sales, Kolon FnC developed innovative
projects in previous years to succeed in the competitive omni-
channel market environment [3]. Its innovations consist of
several information technology projects to attract customers,
establish relationships with customers, and increase online
sales.
Two critical issues in online sales are the high return rates
and consumers’ hesitancy to purchase apparel. Previous
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3023040, IEEE Access
Author Name: Preparation of Papers for IEEE Access (February 2018)
2
research [4] have identified online customers reluctance to
buy apparel because they were not able to wear them to assess
the style. In addition, they could easily return a product if it
happens to be the wrong size or fit. Thus, our initiatives have
focused on providing relevant, detailed product information
and new experiences with 3D virtual try-on to attract more
customers and help them with their decision. Before the
advent of our 3D virtual try-on service, our online customers
only had 2D model pictures and garment size information to
make a purchase decision leaving it to their imagination
weather the apparel will fit. It suits means that the customer
likes the style, and ‘it fits’ means that the garment fits the
customer’s body well. Since online customers are unable to
wear a particular garment and do not know whether it looks
good, they are reluctant to purchase a garment and tend to
return garments for the wrong size and fit. To solve this
problem, our project concentrates on a 3D virtual model with
a try-on and mix-and-match feature. With this feature, online
customers can make a virtual model of their own shape and
size and ‘try on’ garments by looking at images of the outfit.
A customer can then decide whether a garment looks good and
fits well. In addition, a customer can mix and match garments
to coordinate a style.
In this study, we use the term virtual try-on (VTO) instead
of virtual fitting because even if a garment shows fit virtually,
readers might be confused for the following reasons. In the
fashion supply chain, fit could be checked not only during the
sales/retail process but also at the designing process. First,
during the design process, a designer checks the fit to illustrate
his or her design intention aesthetically, and a technical
designer checks the fit to ensure that the garment fits well on
the body and to fix the patterns to construct well. A pattern
professional must then fix the garment patterns according to
the fit guide of the designer and the technical designer. Second,
during the sales/retail process, a customer checks out the fit,
which helps to make a decision on purchasing apparel. The
replacement of physical fitting with virtual fitting would save
the costs of making samples in the design process and help
customers to buy without having to wear an actual garment. In
this study, we deal with virtual fitting during the sales process
and use the term VTO.
Most previous studies on virtual fitting or virtual try-on
(VTO) have focused on technology [5, 6, 7], the effects and
effectiveness [8, 9, 10, 11], and consumer behavior [12, 13,
14]. These were either studied on its technology in design
chain from the point of view of making clothes, or studies on
effects and behavior from the point of view of customers.
However, no research has been conducted to investigate the
effect of 3D virtual try-on on sales with actual data. Therefore,
this study is to examine how 3D virtual try-on affects the sales
results. Then, we explain customers’ experience of 3D virtual
try-on service by interviewing a group of users of a 3D virtual
try-on service. In this study, we explain in detail the advanced
information technology of virtual try-on and analyze the
effects of information technology on sales in terms of the
relationship between customer experience and real sales
results. The purpose of this study is to measure the effects of
customer’s virtual try-on on the actual sales and to explain the
meanings of this customer experience and how it might create
value for online customers. Therefore, we developed the
following research questions:
RQ 1. To what extent is the virtual try-on experience able
to affect sales and consumer behaviors?
RQ 2. What are the values and meanings of virtual try-on?
This study is organized as follows. First, the relevant
literature about virtual fitting is outlined and the gaps from our
study and others are explained. Next, the research design is
explained and the key findings are discussed. Finally, the
implications and limits of our research are discussed.
II. LITERATURE REVIEW
A.
VIRTUAL TRY-ON(VTO) AT RETAIL/SALES CHAIN
Virtual try-on (VTO) is a new technology used to help
customers try on and mix and match apparel without a fitting
room. In the retail environment, U.S. department stores such
as Macy’s, J.C. Penney, and Nordstrom’s applied new virtual
fitting room technology in the early 2010s to improve their
customers’ retail experiences. It appeared when conventional
shops were in crisis because of consumers’ shift to online
retailers.[15] VTO technology has been adopted because it has
many advantages in both retail channels. At traditional
retailers, many consumers feel reluctant to talk about their
body size and shape with the salesperson, and at online
retailers, consumers are unable to wear the apparel physically.
VTO can then provide consumers with impersonal interaction
with a virtual model of the customer’s own body that replaces
the actual try-on, so that it can conveniently provide size and
fit guide. It can solve the problem of incorrect size and fit,
which is a common reason for consumer returns. VTO data are
significant for retailers because they can explain consumer
insights. Because digital simulations can track consumers’
preferences in buying clothes, retailers can use this
information to predict sales. In addition, because VTO tends
to increase consumer satisfaction, retailers can use this tool to
increase consumer loyalty. Recently, consumers can try
clothes on a model more similar with them by using
personalized 3D model using their personal information when
they purchase clothes online [16]. In addition, they can check
the wearing sensation not only from the front but also from the
back and sides with the use of various angles. Despite these
advantages, VTO has shortcomings, including poor
simulation of the tactile sense or texture of the fabric. So far,
VTO has been developed with advanced technology including
Microsoft’s Kinect, 3D scanning technology, physical
simulation with machine learning techniques, GPU-based
real-time cloth simulation, image-based virtual try-on network,
multi-pose guided virtual try-on network, and better size
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3023040, IEEE Access
Author Name: Preparation of Papers for IEEE Access (February 2018)
3
recommendations [8] [17] [18] [19]. Recent technology has
focused on realistic simulation [20]. Sun et al. (2019) proposed
a novel image-based virtual try-on network, which could
maintain the structural consistency between the generated
image and original image by human parsing (see Figure 1)
[19]. Boonbrahm (2015) [5] researched realistic simulation
technology of 3D dress and the physical properties of fabrics
by varying parameters. We can thus tell the kinds of fabrics
used by looking at the appearances of fabric stretching
stiffness, bending stiffness, damping, friction, collision, and
gravity. This simulation is possible with a 3D game engine,
Maya, and Microsoft’s Kinect2.
FIGURE 1. Image-based virtual try-on network [42]
VTO technology consists of making a virtual body model
from the customer’s own body size, 3D garment modeling,
and interactive try-on and mix-and-match of garments.[21]
Many studies explain the method of making a virtual body by
scanning or measuring the customer’s body. 3D garment
modeling is done from 2D computer-aided design (CAD) flat
patterns or from 2D garment photos. Because the creation of
3D garments has been developed using 2D garment photos,
VTO can be easily implemented to build a virtual showroom
for people who do not have 2D CAD patterns.[22] VTO has
clearly begun to revolutionize the retail scene with advanced
technology such as multiple-sensor 3D scanners, augmented
reality, and simulations. Also, making a good virtual fitting
room relied on two important factors, the method of detecting
the user’s body size, position, and movement, and the method
of displaying the virtual garments superimposed in the user’s
body [23]. In addition to VTO, social networking features
allow customers to send photos and receive quick
feedback.[24] According to the research examining VTO’s
results, it helps to boost sales and decrease returns [25].
Previous studies have concentrated on modeling technology
and realistic simulations. They evaluated that VTO technology
overall provides adequate visibility to replace physical try-on
and to check the size and fit for customers. However, rather
less attention has been paid to the effects of VTO in relation
to retail sales.
B.
VIRTUAL FITTING IN THE DESIGN CHAIN
Fitting is the process by which designers and technical
designers check the garment details, size, and fit of the sample
on a model body and seeks a final pattern by iteratively fixing
the sample shape. Virtual fitting is the visualization of a
sample garment to tune in the fits virtually. In a retail
environment, virtual fitting has become an important aspect of
the design process because it can replace some of the samples
needed for fitting and can reduce production costs. As an
example, the U.S manufacturers Target, Kohl’s, and Levi’s
have adopted 3D simulation technology to check fit in the
garment sample-making process because of its benefits of
speed and lower total production costs, and its use is likely to
expand in the near future [26]. Retailers understand the weak
points of virtual fitting, which is insufficient to visualize fits
and garment shapes from the perspectives of designers and
technical designers to replace the physical fitting process, but
several U.S. manufacturers plan to use virtual fitting tools in
sample production for cost reasons [6].
In general, virtual fitting technology is similar to
visualization of the garment, yet it differs from VTO in two
ways. First, virtual fitting is usually deployed on apparel
patternmaking CAD systems to assemble 2D patterns of a
garment to drape on a 3D body to shows how it fits on the
body. Several virtual fitting CAD programs have been
developed by the pattern CAD companies, including Optitex
(Israel), Gerber Scientific (US), Lectra (France), Technoa
(Japan), Assyst (Germany), and CLO Virtual Fashion
(Korea).[27] Because the purpose of virtual fitting is to tune in
and complete the patterns to reach a quality garment, the
technology is developed by pattern CAD companies, and the
3D geometric rendering is from 2D CAD pattern images.
Second, virtual fitting requires more advanced technology
than VTO, because virtual fitting is used to check the fit and
fine tune a garment when making a sample. If virtual fitting
can replace conventional sample fitting during the design
process, it would be advantageous to save sample-making
time and total production cost [28]. For clothing retailers,
virtual fitting has become a favorite technology for investment.
To replace actual fitting, virtual fitting simulation must
illustrate silhouette, detail sizes, fits, ease, realistic fabric
folds/drapes, and fabric color/texture that looks the same as
the actual sample [7, 9]. For example, Dongdaemun provides
stores sponsored by Seoul with a virtual fitting service of
producing clothes from virtual clothes designed by customers
using 3D technology within 24 hours. This service is called
‘Within 24.’ Consumers choose basic design samples (design,
fabric, pattern, etc.) from look books of the 3D cloth
manufacturing software and try the samples on their avatars to
make real clothes.
Even though visualization technology is advanced and
overall adequate to represent fits, current virtual fitting is
generally ineffective and cannot replace a designer/technical
designer’s fitting process of the actual sample. Studies have
investigated the similarities of actual garments and virtual 3D
pants, skirts, jackets and other items. One study of pants
explained that the waist location, ease, and stress folds of
virtual pants differed greatly from actual sample pants.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3023040, IEEE Access
Author Name: Preparation of Papers for IEEE Access (February 2018)
4
Another study concluded that the overall accuracy of virtual
fitting was good but insufficient for designers to replace
physical fitting [6]. Another study evaluated that visualization
of size variations was enough for viewers to perceive size
differences, which means that this size visualization could be
used in the fit analysis of grading samples. As a result,
manufacturers can replace fit analysis of size grading samples
and need not make multiple samples in various size grades or
hire multiple fit models of various sizes [10]. However, virtual
fitting is limited and has issues that must be addressed before
virtual technology can be completely trusted as a visual fit
analysis tool.
Several researchers [11, 29, 30] have evaluated the fit and
style of skirts by comparing similrities between the real fit and
the virtual fit. They found that the accuracy of visualization—
for example in terms of shirring, frill, fit point placement, and
fabric properties—is insufficient for use by designers and
technical designers. In these studies, investigated the fit of
various items and concluded that it was generally possible to
visualize overall fits, but that its limits prevent it from
replacing actual fitting completely. While they recognize the
value of virtual fitting and the new technology of simulation,
they do not think it sufficiently mature to provide precise fits
to replace sample fitting to fix the patterns. They have
recommended appropriate feedback for better visual
representation. This study is limited to VTO, which is a part
of the virtual fitting.
C. VIRTUAL TRY-ON AND CONSUMER BEHAVIOR
Many researchers have studied customers’ attitudes about
virtual fitting features online. Thus far, studies have been
confined to analysis of a psychological model, and rather less
attention has been paid to the effects of VTO on actual sales
results. One study tested the theoretical TAM (technology
acceptance model)[31] and found links among body
satisfaction, the VTO experience, attitude toward the product,
and online purchasing intention. The findings were as follows.
Women with higher body satisfaction tended to perceive more
enjoyment with the VTO feature. Body satisfaction was not
related to the perceived usefulness of the virtual fitting. Both
hedonic and utilitarian aspects of virtual fitting play a role in
determining purchase intention [32]. Another study
investigated the stimulus-organism-response psychological
model. It studied the effects of viewing a virtual version of
one’s own body (or ideal body) and body satisfaction on a
woman’s concern for garment fit/size and attitude to use VTO.
The findings are as follows. Women who were exposed to a
virtual version of their own body measurements showed
greater concerns about garment fit and size than those who
were exposed to an ideal model. It was also found that women
who have (1) greater concerns about garment fit and size and
(2) lower body satisfaction are more likely to adopt VTO [12].
Another study emphasized the hedonic value of image
interactivity of a virtual fitting room and its effects on online
approach responses. It evaluated the effects of image
interactivity (limited to the mix-and-match feature) and found
that it acted as a stimulating experience and predicted
emotional arousal and pleasure to lead, willingness to
purchase, and willingness to patronize the online store. It
means that the use of the mix-and-match feature for
information and for fun is a key to involving emotion and can
result in a positive attitude in online responses [33]. In addition,
other studies have examined VTO as a marketing feature of
mobile shopping and the value of mobile shopping and its
rapid growth. They attempted to illustrate qualitative analyses
of mobile shopping in terms of convenience and experience
aspects. These studies have explained the main drivers of
consumer behavior in a mobile shopping scenario and
illustrated the hedonic and utilitarian value from the use of
new features including virtual fitting [13].
However, even though many studies have reported on VTO
technology and its value, very few studies have reported its
effects on actual sales and fundamental meanings of consumer
behavior. Therefore, the purpose of this study is to understand
and interpret the qualitative meanings related to the consumer
experience with VTO and measure its effects on actual sales.
We define virtual fitting as limited to customers’ try-on at the
value chain of ‘retailer/sales,’ which is not virtual fitting
during the design process. Our research was performed using
the VTO features of L brand, a women’s apparel brand.
III. CREATING VIRTUAL 3D GENERIC BODY MODEL
AND 3D GARMENT DIGITIZATION
The process of 3D general body modeling, cloth simulation,
and customized body modeling is organized as follows (see
Figure 2). First, in ‘1) 3D model creation (human body
modeling)’, we created a virtual 3D model by digitizing the
human body. We analyzed over 1,000 body measurements
from our database through 10 line JavaScript code. Here,
34~54 sizes are divided into 11 different sizes through grading.
In addition, the 3D model can be created with 5 somatotypes
including the standard body shape and 4 different body shapes
(see Figure 3). The somatotypes include five types, including
4 different body shapes and 1 standard body shape. Second, ‘2)
Garment photography’ is a picture of a garment dressed on a
mannequin. It is shot using an automated mannequin handling
system. Since the clothes are photographed in mannequins, the
impression of clothes in real human body can be displayed.
Shooting time is 5-9 minutes per PCS (clothes), saving effort
and time compared to shooting (fitting) a real model offline. It
also does not require a professional photographer. Therefore,
cost-saving effect is expected.
Third, in ‘3) Digitization (creation of clothed 3D model)’, we
completed a clothed model through the digitization and
rendering of the garment. That is, we performed the work of
cutting out the photographed clothes image and rendering it to
the 3D model. These cut outs make the photographed clothes
look more realistic by improving the ability of the rendering
on the model. These cut outs drive our ability to render 3D
models. The virtual 3D modeling and the cloth simulation of
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3023040, IEEE Access
Author Name: Preparation of Papers for IEEE Access (February 2018)
5
this study are a collaborative project with M company, a
London-based tech startup that enables VTO online. Finally,
in '4) QC (quality control) and realistic effect representation (6
motions)’, we check out 3D visuals from multiple body shapes
and every angle. And here, a model of six angles was
completed. That is, it is used to create models of outfitted
images from various angles and to rotate them in try-on. The
try-on and the mix-and-match services are a part of the online
mall feature at L brand, a women’s casual brand.
FIGURE 2. Process of 3D body modeling, cloth simulation, and customized body modeling
A. GENERIC BODY MODELING – 5 SOMATOTYPES FOR
11 SIZES
To make the VTO fun and interactive, our case study focuses
on creating an attractive body model, realistic cloth simulation,
and diverse poses to display the outfit well. We considered
four major issues. (1) How can we create a customized model
with the best quality as we transform the generic body model
into an individualized body based on a customer’s body
measurements? How can we reflect varied body shapes and
body sizes to cater to all customers? (2) What features make
our virtual model more attractive? How can we generate a
realistic body and skin tone? (3) When we simulate a garment,
can we illustrate the size and fit accurately to help the customer?
Can we express the fabric’s physical properties well and
express the drape and wrinkles so that a customer can feel the
texture of the clothes? (4) How can we provide interactive and
realistic simulation when a customer tries-on a garment? What
features make our simulation more fun and more interactive?
Focusing on these four issues, we created 3D generic models
with diverse body shapes and sizes to embrace various
customers’ bodies. We then modified a generic model to an
individualized body to reflect different body characteristics
based on the customer’s measurements.
The first step in virtual female body modeling was to create
a generic female body model from the standard body size
measurements. We used more than 1,000 points of body
measurements from our database. Size classification, or size
grading, was then conducted mainly on the basis of height
(stature) and waist circumference (girth). The sizes category is
divided into 11 sizes from 34 to 54. Basically, the 11 sizes are
identically described as the standard body shape using linear
grading. That is, new dataset of body shape is created using
linear grading as well as the relevant part’s size based on
individuals’ body sizes in order to generate more similar
models with each consumer. The data of 11 sizes was created
with the analysis by using machine learning based on the M
company’s accumulated data. The second step was to build
four more body shapes for each size category because the
standard body type is not applicable to all customers. Our data
reported that only 22% of our 1.6 million customers could use
standard body measurements, so we created four additional
body shapes to cater to all other varied body shapes. When we
defined additional body shapes, we used many specific sets of
measurements from the data cluster analysis, and these body
characteristics differed significantly from the standard shape.
The body shapes are classified as ‘more pear-shaped,’ ‘more
hourglass,’ ‘more rectangular,’ and ‘more inverted triangle’
[34]. In this study, we generated five groups of body shapes
based on the customers’ profile, and we defined the ‘standard
body shape’ as having similar proportions of the
circumference of the bust-waist-hip and defined the other four
somatotypes as follows. As noted, our data indicated that 22%
of our 1.6 million customers have a standard body shape, 45%
are more pear-shaped, 21% have a more hourglass shape, 10%
have a more rectangular shape, and only 2% have a more
inverted triangle shape.
1) Standard body shape—The bust and hips are basically
the same size and similarly proportioned on the
circumference of the bust-waist-hip.
2) More hourglass (rounded)—The bust and hips are larger
than the standard body shape, and the waist is well
defined.
3) More pear-shaped (triangular)—The hips are large, and
the bust is narrow. The hips are fuller than those of the
hourglass body silhouette, and the bust is narrower than
that of the hourglass silhouette.
4) More rectangular (straight)—The bust and hips are
narrower or wider than those of the standard body shape.
5) More inverted triangle—The bust is large, and the hips
are narrow. The bust is fuller than that of the hourglass
body silhouette, and the hip is narrower than that of the
3D model creation Garment
photography Digitization QC & Realistic
effect representation
BODIES
GARMENTS
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3023040, IEEE Access
Author Name: Preparation of Papers for IEEE Access (February 2018)
6
hourglass silhouette.
The third step was to make our female model more
attractive and realistic using the techniques of animated pose,
color expression, and realistic skin texture. We built various
hair styles, facial tones/expressions, and stylish underwear for
the virtual model. We tried to represent realistic skin to bring
in vividness and built a variety of poses. We tried to present
live skin and build six other poses at every 60° angle rotating
view for being realistic. Figure 3 shows the creation of a 3D
model that reflects 11 body sizes and 5 body shapes based on
consumer measurements. That is, when a consumer enters
his/her measurements, 3D virtual try-on uses the consumer’s
size and body shape type to create a 3D virtual model. The
body shape types included the different body shapes in
addition to the standard body shape. The images of standard
body shape and different body shapes are listed from the left
side, and the standard body shape of green body curve is also
shown on each different body shape image to distinguish
different body shapes by body shape. In other words, an
individual’s body shape is generated among five body shape
groups based on the individual’s own body sizes, based on
which, 3D model including every angle is developed. Figure
4 shows 3D virtual models with different facial
tones/expressions, hair styles, poses, etc.
FIGURE 3. Process of creating the 3D model with 5 body shapes and 11 sizes
FIGURE 4. 3D virtual models
B. GARMENT DIGITIZATION AND CLOTHED BODY
MODELING
Clothed body modeling is used to create a 3D clothed body
model by rendering a 3D garment mesh surface on the body.
We propose a new method, instead of 2D CAD patterns, we
used 3D garment photos; to make a 3D garment mesh surface
because 2D CAD patterns are always copyrighted and require
professional CAD software, therefore our study propose an
efficient way. This process begins with 2D garment
photographs taken with an automated photo studio system.
The photographs of garments on mannequins were taken with
customized equipment. Since the clothes are photographed in
mannequins, the impression of clothes in real human body can
be displayed. Our system does not require a professional
photographer, and because one coordinator is wearing a
mannequin costume, it takes only 5 to 9 minutes to photograph
a single piece of clothes, and up to 200 photos can be taken
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3023040, IEEE Access
Author Name: Preparation of Papers for IEEE Access (February 2018)
7
each day. It thus saves time and money compared to
conventional photographing methods.
From the 2D garment photographs, the first step is to
remove the background and extract the garment boundary
using image processing techniques such as edge detection and
extraction of feature points. To find a feature point to build a
garment silhouette, each pixel in each column/row is
examined to find the first and last background color for each
column/row. The garment region is then extracted. Among the
candidate feature points outlined here, we used an algorithm
that selects a feature point of high curvature using a predefined
feature point mask. We used a supervised learning algorithm
(scaled conjugate gradient) and extended it with new options
[35, 36].
The second step is to reconstruct the garment surface inside
the garment silhouette. To make its surface smooth and
balanced, we use an algorithm to insert the interior nodes
uniformly inside the garment silhouette. Delaunay
triangulation is applied to produce structured triangles by
connecting points to form a convex polygon. In addition, some
new nodes are inserted, and some overlapping nodes are
eliminated to make the initial garment surface smooth, and
Delaunay triangulation is again applied to distribute those
nodes evenly. This process is iterated, and triangles whose
edge is outside the garment region are removed for a good
finish.
After the garment surface is created, a voxel-like process is
applied to render the planar garment surface mesh onto the
virtual body. The garment’s triangulated surface is then placed
near the virtual body, where feature points are used. Between
the body and the 3D garment, an efficient collision
detection/response algorithm is used to move the 3D garment
around the body model. We then sewed eight pieces of surface
views at every 45 degrees with a feature matching method and
used an algorithm to minimize the stitching seam to create a
seamless texture. The processing time for 3D garment
digitization of each item from the garment mesh is less than 1
second. Figure 5 represents the digitalization of clothing and
the process of clothed body modeling.
FIGURE 5. Digitalization of clothing and clothed body modeling
IV. CASE STUDY
A. CUSTOMIZED BODY MODEL AND VIRTUAL TRY-ON
Our VTO study was conducted with the online mall and
mobile application of L brand, a women’s wear brand. The
most important thing is that the customer enters only five
measurements, but our customized model reflects the
individual’s body shape and looks attractive with the target
garments. To create attractive individualized models, we
provided 25 models with various hairstyles and facial
expressions (see Figure 2 and Figure 3). We also focused on
accurate size and fitting with speed. The process for this VTO
included four major steps. When a customer enters the main
page, she selects the ‘try’ icon on the left corner of each
fashion style image. The fashion styles are categorized by
brand, type, size, and color to allow the customer to filter
favorable items. The construction of an individual model then
begins. The customer enters five body measurements: height,
weight, bra size, body shape, and leg length category. If she
wishes a more elaborate body shape, she can enter three more
body measurements: waist girth, hip girth, and inner leg length
(i.e., from the crotch to the inner ankle bone). She then chooses
her own model from among 25 faces and hairstyles. With the
customer’s measurements, the generic model is first modified
into a customized model.
The customer can now view a virtual adorned model and
rotate it to six viewing angles. Our service recommends the
right garment size and a fit guide, which includes tighter,
comfortable, and looser around the waist and hips. A customer
also can select other items to mix and match and search for
coordinated items. Finally, the customer can share a
photograph of the outfit model via SNS and obtain feedback
and opinions from others. This provides the customer with a
new retail experience.
3D MODELING
GARMENT
PHOTOGRAPHY
•Voxel
creation
•Background
removal
•Feature points
detection
GARMENT
DIGITIZATION
Cloth Simulation
System
Clothed Body Model
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3023040, IEEE Access
Author Name: Preparation of Papers for IEEE Access (February 2018)
8
B. QUALITATIVE INTERVIEWS ABOUT VIRTUAL TRY-
ON
Our VTO service was carried out online in the mobile
shopping mall of L brand from June 1 to August 13, 2017. A
total of 105 SKUs of 2017 S/S styles were presented, and 8058
customers used the VTO service. We conducted a telephone
interview with consumers who used the service to get
feedback on the virtual try-on. In addition, we collected
feedback about the service on our website and blog. In the
telephone interview, we encouraged the consumers to talk
freely about their experience and feedback about purchasing
clothes using virtual try-on. It took about 15 minutes per
person for an interview. We recorded the interviews and later
transcribed them. Then the significant phenomena within the
data extracted from the interviews were conceptualized and
coded to categorize similar or related experience. In addition,
we web-crawled the reviews on using virtual try-on through
the website and blog to use them in the analysis. We used open
coding to analyze the transcripts. Open coding is the initial
step of theoretical analysis that pertains to the initial discovery
of categories and their properties [37, 38].
During the open coding, each coder examined the interview
responses and the comments line by line to identify codes
within the textual data that could explain customers’
experience. The coder then discussed each concept identified
and named it after arriving at a consensus. We grouped these
concepts into broader categories that reflected commonalities
among the codes. Thus, we used open coding to delineate
concepts and group them into meaningful categories. To
validate the data coding results, three other researchers
reviewed the interview details and keyword classifications.
We derived concepts from the open coding of the
qualitative approach about consumers’ virtual try-on
experience of this study. As a result, we extracted six main
concepts: confidence in apparel fit/style, model self-congruity,
convenience, model styles, pleasure, and purchase intention.
The result of the analysis showed that respondents mentioned
the most about ‘confidence in apparel fit/style (32.2%).’ Due
to not being able to see the suitability on online mall it raises
concerns for consumers; however, virtual try-on in this study
recommends size that fits well with 3D models similar to
consumers satisfying their concerns. The actual responses
include “The apparel will fit right, and the size
recommendation service is nice,” “I am pleased with the look
of the right length pants/skirt.” In addition, they said they
wanted more detailed description about body size required in
developing 3D model for virtual try-on.
Next, they mentioned a lot about ‘convenience’. Consumers
were bothered with having to change clothes several times
when buying clothes. However, they mentioned that virtual
try-on was convenient because they did not have to wear
clothes. In addition, the use of a virtual try-on system was
found to be convenient. The actual answers include "It's
convenient because I don't have to wear clothes many times,”
“It's easy to dress the 3D model. In addition, ‘model self-
congruity’, ‘model styles’, ‘pleasure’, and ‘purchase intention’
were mentioned a lot in this order. With regard to ‘model self-
congruity’, thinking a 3D model created by the virtual try-on
as themselves, they wanted to make the model more similar or
better than themselves. With regard to ‘model styles’, the
contents related to quality or styling of the 3D model were
mentioned, and with respect to pleasure, the use of virtual try-
ons was found to be interesting. They also said they wanted to
change some parts to become more similar with them in
developing 3D model. That is, they can choose clothes that fit
better if 3D model is developed based not only on body size
but also on the face shape and hair style. Specifically, while
they were satisfied with the current 3D model, they wanted the
face shape similar to them and realistic simulation as if they
tried on real clothes. Finally, with regard to ‘purchase
intention’, it means that consumers are willing to purchase
clothes through virtual try-on. The actual answers include
“The next time I buy clothes,” “I will use virtual try-on to buy
clothes” etc.
TABLE I
CODING RESULTS: CUSTOMER RESPONSES
No
Customer response
Comment examples
1
Confidence in
apparel fit/style
The apparel will fit right, and the size
recommendation service is nice.
I am pleased to look at the right length
of pants/skirt.
I want a more detailed description
about the size/fit.
2
Model self-
congruity
The model I used to try-on clothes is
consistent with how I see myself.
The model looks like me, but if I could
change the avatar, I would like to
change its proportion for a better look.
3
Convenience
The try-on service is easy to dress the
3D model.
It is convenient because I do not have
to wear clothes many times.
4
Model styles
I am satisfied, but I want to view an
avatar with my face.
I hope it provides the realistic
simulation of cloth.
5
Pleasure
It is fun and awesome
6
Purchase intentions
The next time I buy clothes, I will use
virtual try-on to buy clothes.
C. EFFECTS OF VIRTUAL TRY-ON ON ACTUAL SALES
RESULTS
We analyzed the effects of VTO on actual sales results. To
measure the effects, we collected customer data and sales data
from June 1, 2017 to August 13, 2017. We then cleaned the
data and filtered the return and exchange sales data to extract
net sales and consolidated the data into one file of customer
net sales data. We analyzed 11,029 transactions, and 8,058
customers used the VTO service. Primarily, we attempted to
validate the effects of VTO in the dimensions of time and
money and compared the sales results of VTO users and VTO
non-users. We measured several indices as follows.
TABLE II
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3023040, IEEE Access
Author Name: Preparation of Papers for IEEE Access (February 2018)
9
INDICES TO ESTIMATE EFFECTS OF VTO
Measures
Definition
Formula
Visitor growth
rate
Growth rate between the
average annual visitors
and visitors during the
VTO service
[(visitor count VTO –
average annual visitor
count) / average annual
visitor count] * 100
Revisit growth
rate
Growth rate from the
average annual revisit
consumer count to
consumer count during
the VTO service
[(Revisit count VTO –
average annual revisit
count) / average annual
revisit count] * 100
Time spent
online
Amount of time a
customer spent online at
every visit
Time spent
Time spent
online growth
rate
Growth rate from the
average annual amount
of time spent to amount
of time during the VTO
service
[(Time spent VTO –
average time spent) /
average time spent] *
100
Virtual try-on
participation
rate
Ratio of the number of
visitors and try-on users
(VTO user count / total
visitor count) * 100
Purchase
conversion
rate
Rate of conversion, total
resulting number of sales
from the number of
entities, business that
potentially attracts
purchase
(VTO items purchase
count / items purchase
count) * 100
Average
number of
purchases per
customer
A count of items a
customer purchases at
one shopping visit
User : (VTO purchase
item count / VTO user
count)
Nonuser : (VTO nonuser
purchase item count /
VTO nonuser count)
Average sales
per customer
Amount of purchase per
a customer
User : (VTO purchase
amount / VTO user
count)
Nonuser : (VTO nonuser
purchase amount / VTO
nonuser count)
Return rate
the rate of return counts
from the sales counts
Return count during
VTO / Total sales count
during VTO
Our analysis illustrates that VTO has positive effects on
visitor numbers, time spent, money spent, and sales amounts
as follows (see Table 3).
TABLE Ⅲ
ANALYSIS RESULTS
Division
Results
Visitor numbers
The revisit rate is increased by 2.24 times after
the use of the VTO (growth rate of 324%).
Time spent
The time stayed online site with the VTO service
is increased by 2.37 time (growth rate of 337%).
Money spent
The average sales per customer is increased by
₩14,000 ($13) after the use of the VTO.
Sales amount
The purchase conversion rate is increased by 2.8
times after the use of the VTO.
The return rate is decreased by 27% after the use
of the VTO.
It is found that visitor numbers, time spent, money spent,
and sales amounts increased through VTO in this study. This
indicates that consumers revisit the site because they are
satisfied with the virtual try-on service as mentioned in the
previous studies that state if consumers are satisfied with a
service, they revisit the site that provides the service [39] [40].
The previous studies [41] [42] also view longer stay on online
or offline malls as a positive phenomenon. This means that not
only are consumers satisfied with the virtual try-on service, but
also they are interested in making 3D model similar with them,
so they use the service while staying longer on the site. Also,
it can be interpreted that they buy clothes easier by actually
dressing a 3D model with a body that is similar to their body
and giving them the same effect as wearing real clothes. The
most key finding is the remarkable 27% decrease in the return
rate, which resonated with other researchers’ opinions. In
online malls, consumers cannot actually try on and buy clothes,
so the return rate is relatively high because they purchase
clothes without actually checking their fit. However, the VTO
in this study allows consumers to dress a 3D model with the
most similar body to their body through a total of 5
somatotypes including 1 standard body shape and 4 different
body shapes as well as 11 sizes and buy clothes, significantly
reducing the return rate. That is, returns due to the wrong
size/fit seem to be decreased with VTO.
V. CONCLUSIONS
This study has three key findings through developing virtual
try-on and examining effects of the VTO on sales results. First,
in this study, we created a virtual 3D model reflecting the more
detailed size and type is generated based on the customer
measurement, thus making it more realistic and natural
looking. Second, customers responded positively to this
service. The most significant finding is that the return rate
decreased by 27% through the service. We used the
consolidated customer net sales data from Lucky Chouette, a
line of women’s casual wear from Kolon FnC, from June 1 to
August 13, 2017. A total of 105 SKUs of 2017 S/S styles were
examined, and 8,058 customers used the VTO service. During
this 2.5 month period, 11,029 transactions took place.
Third, our observations of customer responses indicated that
they were satisfied with the VTO service with various values.
The marketing and e-commerce literature consider customer
value to be an important predictor of customer behavior and
decision-making [43, 44]. Customer value has been used as a
theoretical foundation in explaining customers’ decisions in
making purchases [45]. In virtual try-on of this study, when
consumers enter their basic body size, a 3D model similar to
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3023040, IEEE Access
Author Name: Preparation of Papers for IEEE Access (February 2018)
10
the consumer is automatically generated, allowing them to
dress the model with clothes they wish to purchase. This
belongs to a functional value among customer value types.
Through the VTO, customers could check the size and fit of
apparel well, especially pants and skirts. For convenience,
filtering the wrong size/fit could reduce customer returns, and
it was proven to result in a 27% reduction. In terms of hedonic
value, consumers feel that it is fun to create a 3D model similar
to them and dress it, which makes them stay longer at the site.
Also, the results show that the average sales per customer
increased by 14,000 won, or $13. To sum up, VTO could
provide convenience, enjoyment, and a new experience that
has a positive effect on purchasing; we conclude that it could
boost purchase intention and sales in the near future. The most
important contribution of VTO is to replace physical fitting
rooms and help retailers reduce their return rates.
One limitation of this study is that only 8,058 customers used
VTO over a period of 2.5 months. In addition, we could not
analyze the mass customer responses. Therefore, if more
customers’ responses are used over a longer period of time for
more items, the results will reflect more general opinions and
elicit better insights with pros and cons about this service.
However, this study has several implications for researchers
and practitioners. First, this study contributes to our
understanding of the entire relationship of VTO, actual sales,
and customer experience and estimates the actual sales effects
via data mining. Second, this study contributes to IS research
by using customer data to study the effects of a virtual try-on
service on sales results. The previous research lacked
examining effects of a virtual try-on on the sales with
customers’ actual data. Third, in this research, the VTO
service customers were interviewed based on the qualitative
research as a way to extract the concepts for using the VTO
service. These extracted concepts were then applied to the
customer value theory to interpret and explain customers’
experience. This study also has practical implications for using
the VTO services. As the interest in the virtual try-on service
increases, many fashion companies try the service. This study
can provide guidelines for companies that expand their
business to the online commerce and companies that plan and
operate the virtual try-on service. Due to the tendency of
reluctance to face-to-face contact due to COVID-19 along
with the development of information technology, the online
commerce activities of untact services are boosted. According
to this tendency and the consumption trend, this study not only
enhances the convenience and satisfaction level of consumers
but also contributes to profit increase of online retailers.
REFERENCES
[1] Marketing Chart. (2018, 2018-12-07). The too most popular
categories purchased online are…. Available:
https://www.marketingcharts.com/industries/retail-and-e-
commerce-106597
[2] Smallbizgenius (2019, 2019-08-02). 40 eCommerce statistics
every retailer should know in 2019. Available:
https://www.smallbizgenius.net/by-the-numbers/ecommerce-
statistics/.
[3] Edaily (2017-09-12). Kolon Fnc, opening of 'omni center'
combined with online and offline stores. Available:
https://www.edaily.co.kr/news/read?newsId=02318966616059
464&mediaCodeNo=
[4] D. Protopsaltou, C. Luible, M. Arevalo, and N. Magnenat-
Thalmann, “A body and garment creation method for an Internet
based virtual fitting room,” in Advances in Modelling, Animation
and Rendering: Springer, 2002, pp. 105-122.
[5] P. Boonbrahm, C. Kaewrat, and S. Boonbrahm, “Realistic
simulation in virtual fitting room using physical properties of
fabrics,” Procedia Computer Science, vol. 75, pp. 12-16, 2015.
[6] H. K. Song and S. P. Ashdown, “Investigation of the validity of
3-D virtual fitting for pants,” Clothing and Textiles Research
Journal, vol. 33, no. 4, pp. 314-330, 2015.
[7] O. Sabina, F. Emilia, A. Manuela, M. Alexandra, P. Georgeta,
and S. Adrian, “Applied 3D virtual try-on for bodies with
atypical characteristics,” Procedia Engineering, vol. 100, pp.
672-681, 2015.
[8] E. Shein, “Computing what fits,” Communications of the ACM,
vol. 57, no. 11, pp. 16-19, 2014.
[9] O. Sabina, S. Elena, F. Emilia, and S. Adrian, “Virtual fitting–
innovative technology for customize clothing design,” Procedia
Engineering, vol. 69, pp. 555-564, 2014.
[10] D.-E. Kim, “Psychophysical testing of garment size variation
using three-dimensional virtual try-on technology,” Textile
Research Journal, vol. 86, no. 4, pp. 365-379, 2016.
[11] A. Porterfield and T. A. Lamar, “Examining the effectiveness of
virtual fitting with 3D garment simulation,” International
Journal of Fashion Design, Technology and Education, vol. 10,
no, 3, pp. 320-330, 2017.
[12] E. Shin and F. Baytar, “Apparel fit and size concerns and
intentions to use virtual try-on: Impacts of body satisfaction and
images of models’ bodies,” Clothing and Textiles Research
Journal, vol. 32, no. 1, pp. 20-33, 2014.
[13] E. Pantano and C.-V. Priporas, “The effect of mobile retailing on
consumers’ purchasing experiences: A dynamic perspective,”
Computers in Human Behavior, vol. 61, pp. 548-555, 2016.
[14] M. Beck and D. Crié, “I virtually try it… I want it! Virtual Fitting
Room: A tool to increase on-line and off-line exploratory
behavior, patronage and purchase intentions,” Journal of
Retailing and Consumer Services, vol. 40, pp. 279-286, 2018.
[15] A. Kramer. (2011, 2018-06-19). The Virtual Fitting Room.
Available: https://www.strategy-
business.com/article/00073?gko=790d3
[16] T. Zhang, L. Cao and W. Y. Wang, " The impact of virtual try-
on image interaction technology on online shoppers' purchase
decision," In Proceedings of the 8th International Conference on
E-Education, E-Business, E-Management and E-Learning, pp. 6-
10, 2017.
[17] H. Dong, X. Liang, X. Shen, B. Wang, H. Lai, J. Zhu,… and J.
Yin, " Towards multi-pose guided virtual try-on network," In
Proceedings of the IEEE International Conference on Computer
Vision, pp. 9026-9035.
[18] T. Su, Y. Zhang, Y. Yu and S. Du, " GPU-based real-time cloth
simulation for virtual try-on," In Proceedings of the 26th Pacific
Conference on Computer Graphics and Applications: Posters,
Eurographics Association, pp. 1-2.
[19] F. Sun, J. Guo, Z. Su and C. Gao, "). Image-Based Virtual Try-
on Network with Structural Coherence," In 2019 IEEE
International Conference on Image Processing, pp. 519-523.
[20] P. Volino and N. M. Thalmann, “Implementing fast cloth
simulation with collision response,” in Computer Graphics
International, 2000. Proceedings, 2000, pp. 257-266: IEEE.
[21] M. Yuan, I. R. Khan, F. Farbiz, S. Yao, A. Niswar, and M.-H.
Foo, “A mixed reality virtual clothes try-on system,” IEEE
Transactions on Multimedia, vol. 15, no. 8, pp. 1958-1968, 2013.
[22] D. Shin and Y. Chen, “3D garment digitisation for virtual
wardrobe using a commodity depth sensor,” in International
Conference on Computer Vision: ICCV 2017, 2017: IEEE.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3023040, IEEE Access
Author Name: Preparation of Papers for IEEE Access (February 2018)
11
[23] C. Kaewrat and P. Boonbrahm, “A survey for a virtual fitting
room by a mixed reality technology,” Walailak Journal of
Science and Technology, vol. 14, no. 10, pp. 759-767, 2017.
[24] C.-I. Cheng, D. S.-M. Liu, C.-H. Tsai, and L.-T. Chen, “A 3D
virtual show room for online apparel retail shop,” in Proceedings:
APSIPA ASC 2009: Asia-Pacific Signal and Information
Processing Association, 2009 Annual Summit and Conference,
2009, pp. 193-199: Asia-Pacific Signal and Information
Processing Association, 2009 Annual Summit and Conference,
International Organizing Committee.
[25] I. Pachoulakis and K. Kapetanakis, “Augmented reality
platforms for virtual fitting rooms,” The International Journal of
Multimedia & Its Applications, vol. 4, no. 4, p. 35, 2012.
[26] K. Salmon, “The coming revolution in retail: Courtesy of 3D
technology,” Kurt Salmon, 2014.
[27] M.-J. Lee and H.-S. Sohn, “A study on the cases of the
application of 3D apparel CAD system to the domestic and
overseas fashion education,” Journal of the Korean Society of
Clothing and Textiles, vol. 35, no. 9, pp. 1112-1124, 2011.
[28] C. Morais and G. Montagna, “Customized wardrobe: Clothing
according to user,” Procedia Manufacturing, vol. 3, pp. 5814-
5821, 2015.
[29] I. Kang and S.-Y. Lee, “A study on the shape of shirring using
3D virtual clothing system,” Journal of the Korean Society of
Clothing and Textiles, vol. 34, no. 7, pp. 1111-1125, 2010.
[30] M.-R. Koo and M.-A. Suh, “A study on the shape of hem-line of
semi-flare skirts according to a cutting angle-Based on the
comparison between real clothing and 3D virtual clothing,” The
Research Journal of the Costume Culture, vol. 17, no. 3, pp. 499-
511, 2009.
[31] F. D. Davis, “Perceived usefulness, perceived ease of use, and
user acceptance of information technology,” MIS quarterly, pp.
319-340, 1989.
[32] U.-J. Yu and M. L. Damhorst, “Body satisfaction as antecedent
to virtual product experience in an online apparel shopping
context,” Clothing and Textiles Research Journal, vol. 33, no. 1,
pp. 3-18, 2015.
[33] A. M. Fiore, H. J. Jin, and J. Kim, “For fun and profit: Hedonic
value from image interactivity and responses toward an online
store,” Psychology & Marketing, vol. 22, no. 8, pp. 669-694,
2005.
[34] G. Pisut and L. Jo Connell, “Fit preferences of female consumers
in the USA,” Journal of Fashion Marketing and Management:
An International Journal, vol. 11, no. 3, pp. 366-379, 2017.
[35] M. F. Møller, "A scaled conjugate gradient algorithm for fast
supervised learning," Neural Networks, vol. 6, no. 4, pp. 525-
533, 1993.
[36] Y. Chen, D. P. Robertson, and R. Cipolla, A Practical System for
Modelling Body Shapes from Single View Measurements. In
BMVC (pp. 1-11), 2011.
[37] J. M. Corbin and A. Strauss, "Grounded theory research:
Procedures, canons, and evaluative criteria," Qualitative
sociology, vol. 13, no. 1, pp. 3-21, 1990.
[38] B. G. Glaser, Basics of grounded theory analysis: Emergence vs
forcing, Sociology press, 1992.
[39] S. Campo-Martínez, J. B. Garau-Vadell and M. P. Martínez-Ruiz,
"Factors influencing repeat visits to a destination: The influence
of group composition," Tourism Management, vol. 31, no. 6, pp.
862-870, 2010.
[40] X. Yan, J. Wang and M. Chau, " Customer revisit intention to
restaurants: Evidence from online reviews," Information
Systems Frontiers, vol. 17, no. 3, pp. 645-657, 2015.
[41] C. Astono, " The effect of shopping mall’s attributes toward
customer satisfaction of ABC mall and XYZ mall," vol. 2, no. 2,
2014.
[42] T. Roux and T. Maree, " Shopper’s Experience of Digital Mall
Signage as Atmospheric Stimuli: An Abstract," In Academy of
Marketing Science World Marketing Congress, pp. 321-322,
2018.
[43] J. N. Sheth, B. I. Newman, and B. L. Gross, "Why we buy what
we buy: A theory of consumption values," Journal of Business
Research, vol. 22, no. 2, pp. 159-170, 1991.
[44] J. C. Sweeney and G. N. Soutar, "Consumer perceived value:
The development of a multiple item scale," Journal of Retailing,
vol. 77, no. 2, pp. 203-220, 2001.
[45] H. W. Kim and S. Gupta, "Investigating Customer Resistance to
Change in Transaction Relationship with an I nternet Vendor,"
Psychology & Marketing, vol. 29, no. 4, pp. 257-269, 2012.
Hyunwoo Hwangbo received his Ph.D in
information systems from the Graduate School of
Information at Yonsei University, Korea in 2017.
He received his Master’s degree in public
administration from Yonsei University, Korea in
2008.
He is an Assistant Professor in the Department
of Global IT Business at Hannam University,
Korea. He was an Adjunct Professor in the
Graduate School of Information at Yonsei
University, Korea since 2019. He was a Senior
Research Fellow of Kolon Benit, co. Ltd., Korea between 2009 and 2018.
He has published in various journals, including Electronic Commerce
Research and Applications, Expert Systems with Applications, Sensors and
the Journal of the Textile Institute. His research interests include data
science, Internet of things and big data processing.
Eun Hie Kim received her B.S, M.S and Ph.D
degrees in clothing and textiles science from Seoul
National University, Seoul, Korea in 2002, her
AAS degree in fashion buying and merchandising
from FIT, NYC, USA in 1995 and her M.E degree
in SCM engineering from MIT, Cambridge, USA
in 2009.
She has been a Big Data Analyst with Kolon
Benit in Korea since 2016. Prior to that, she was
SCM deputy manager at LG fashion from 2013 to
2016, an SCM consultant at Deloitte from 2012 to
2013, a journalist at Fashionbiz from 2003 to 2008 and a full-time lecturer
at universities in Korea from 1997 to 2002. Her interests include fashion
trend analysis, demand forecasting, S&OP, logistics, product lifecycle
management and AI in the fashion industry.
So-Hyun Lee is an Associate Professor in the
School of Management at Xi'an Jiaotong
University, China. Before joining Xi'an Jiaotong
University, she was a Research Professor in the
Graduate School of Information at Yonsei
University, Korea. She received her Ph.D degree
in Information Systems from the Graduate
School of Information at Yonsei University,
Korea in 2016. Her research focuses on digital
business, online marketing, and text/data
analytics. Her research work has appeared in Information Systems Research,
Communications in the ACM, Information & Management, International
Journal of Information Management, Behaviour & Information Technology,
Journal of Database Management, and Internet Research.
Young Jae Jang received his Ph.D. degree in
mechanical engineering from Massachusetts
Institute of Technology (MIT) in 2007. His current
research includes the "Smart Factory" and
"Intelligent Supply Chain and Logistics System."
Hi is cureently the Director of the <Shinsung-
KAIST AI-based Automated Material Handling
System Research Center> leading the multi-
million dollar industry collaborative research
project with the industry partner, Shinsung FA, Inc,
a global factory automation system provider. He has been also involved in
the KAIST On-Line Electric Vehicle (OLEV) project to develop and
commercialize the innovative wireless charging electric vehicle. The project
was recognized as the``50 Best Innovations of 2010'' by TIME Magazine.
His role in the project is to develop the optimal energy management system
to integrate the vehicle system to the road traffic network. He has published
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3023040, IEEE Access
Author Name: Preparation of Papers for IEEE Access (February 2018)
12
numerous technical papers on OLEV technology. Before he joined KAIST,
he worked at Micron Technology, Inc., VA, USA, where as project manager,
he led a global initiative to improve the efficiency of the manufacturing
facilities located worldwide.
