You are on page 1of 17

Article

Clothing and Textiles


Research Journal

Investigation of the Validity of 2015, Vol. 33(4) 314-330


ª The Author(s) 2015
Reprints and permission:
3-D Virtual Fitting for Pants sagepub.com/journalsPermissions.nav
DOI: 10.1177/0887302X15592472
ctr.sagepub.com

Hwa Kyung Song1 and Susan P. Ashdown2

Abstract
The effectiveness of 3D virtual fitting technology when visualizing the fit and silhouette of pants by
analyzing the similarities between real and virtual fit using 20 fit locations, 3 lower body shapes, and
fit status was investigated. We produced 61 custom pants for the participants and virtually tried each
on a personalized 3D body scan avatar. The technology was not generally effective for visualizing
pant fit. Especially, the waist placement of the virtual pants was lower than that for the actual pants.
The virtual software indicated less ease than the actual pants and could not express stress folds due
to slight misfit. The front silhouette of the virtual pants spread wider than the real pants. The virtual
pants for females with a hip tilt shape had the greatest divergence from the actual pants. The virtual
pants with good fit appeared more effectively than did those with poor fit.

Keywords
3-D virtual fitting technology, pants, body shape, fit, body scan, wearable technology

Currently, several retailers, including Target, Kohl’s, and Levi’s, are successfully piloting the use of
three-dimensional (3-D) simulation technology to produce their garment samples (Salmon, 2014;
Yoon, 2013). The clear benefits demonstrated by early adopters of this concept are that 3-D technol-
ogy offers both retailers and manufacturers ‘‘speed’’ and ‘‘lower overall costs.’’ According to a
report by Salmon (2014), 3-D virtual fitting technology lets apparel manufacturers produce far fewer
physical samples for each garment beforehand, decreasing to two samples from the average of three
to five that were needed formerly. Further, this technology can cut the total amount of time needed
for designing and prototyping by an estimated 50%.
Such 3-D simulation technology also improves the communication between the designers, pat-
ternmakers, and technical designers (Salmon, 2014; Yoon, 2013). With most production now
occurring offshore and fit sessions taking place in multiple locations worldwide, it has been dif-
ficult to create conditions that produce a consistent fit from location to location and to communi-
cate any problems and inconsistencies of a prototype back to the vendor producing the garment
(Song & Ashdown, 2010). Currently, when samples are produced offshore, designers and

1
Department of Clothing and Textiles, Kyung Hee University, Seoul, Republic of Korea
2
Department of Fiber Science and Apparel Design, Cornell University, Ithaca, NY, USA

Corresponding Author:
Hwa Kyung Song, Department of Clothing and Textiles, Kyung Hee University, Seoul 130-701, Republic of Korea.
Email: hksong@khu.ac.kr

Downloaded from ctr.sagepub.com by guest on September 8, 2015


Song and Ashdown 315

technical designers communicate on a sample fit using only pictures of the samples on fit models
or dress forms in their tech pack; according to Virginia Meckley, Manager of Technical Design at
Victoria’s Secret, one of the biggest challenges in the apparel business is the lack of accurate
visual tools (Speer, 2008).
3-D scans and 3-D simulation technologies can be valuable, potential tools for visual fit analysis.
3-D images of the clothed figure can be zoomed and rotated, making it possible to comprehend the
depth, origin, and path of stress folds on the entire body (Song & Ashdown, 2010). Although 3-D
technologies have been adopted by only large companies and on a relatively small scale for design-
ing and prototyping, retailers are starting to understand the implications of 3-D technology, given
how impactful the benefits of the technology can be for the entire industry (Salmon, 2014). If the
accuracy of this 3-D simulation technology is demonstrated, more apparel retailers will opt to use
these technologies more often (Apeagyei & Otieno, 2007).
Several researchers have investigated the accuracy of the 3-D garment simulation technology for
visualizing fit. Although Eccles (2011) indicated that pants are the most challenging clothing item
for good fit, few researchers have verified the virtual fit information and 3-D visualization process
for pants. Lee, Nam, Cui, and Choi (2007) analyzed the degree of similarity between real and virtual
fit, while Kim and LaBat (2013) evaluated the accuracy of the 3-D simulation technology in an
online shopping scenario from the customer’s perspective.
The researchers in this study investigate whether the 3-D virtual fitting technology can effec-
tively visualize the fit and silhouette of pants by analyzing the following three key factors: lower
body shapes, various fit statuses, and fit locations using the fashion designers’ or technical
designers’ perspectives. The results of this study provide greater understanding of precisely how
fashion and technical designers evaluate the accuracy of 3-D virtual fit technology for a better
visual fit analysis. The study also provides valuable new information for 3-D virtual fitting soft-
ware developers.

Literature Review
3-D Applications for Garment Simulation
The first 3-D application for cloth simulation in the field of computer graphics appeared in the early
1990s (Carignan, Yang, Magnenat-Thalmann, & Thalmann, 1992; Lafleur, Magnenat-Thalmann, &
Thalmann, 1991). These early virtual representations used approximation and simplification meth-
ods. For instance, they simply simulated the drape of a flag or a rectangular cloth over a typical geo-
metric shape (Volino, Cordier, & Magnenat-Thalmann, 2005). With advances in technology from
the first decade of the 21st century, such as the dynamic manipulation techniques for cloth mesh and
real-time garment prototyping, 3-D applications for garment simulation evolved further and are now
developed enough for the apparel industry to utilize them successfully for virtual garment design and
prototyping.
Two cloth simulation systems are currently available. The first provides general cloth simulation
and animation and is used to design garments. These systems are usually plug-ins for 3-D computer-
aided design (CAD) programs. For instance, MayaCloth is integrated into Maya, Reactor, Stitch, and
SimCloth, while ClothReyes is integrated into 3-D Studio Max, and Dynamics is integrated into
Cinema 4D (Volino et al., 2005).
The second system is used for two-dimensional (2-D) apparel pattern assembly and draping
on 3-D virtual mannequins for visualization or prototyping of garments still in the production
stage. Several of these 3-D virtual fitting CAD programs have been developed by apparel pat-
ternmaking CAD companies, namely, 3-D Runway Creator for Pattern Design Software (PDS)
by Optitex (Israel), V-Stitcher by Gerber Scientific (USA), 3-D Fit by Lectra (France), i-designer

Downloaded from ctr.sagepub.com by guest on September 8, 2015


316 Clothing and Textiles Research Journal 33(4)

by Technoa (Japan), Vidya by Assyst (Germany), and CLO 3-D by CLO Virtual Fashion
(Korea; Lee & Sohn, 2011). These programs usually operate as follows: (a) either a parametric
virtual mannequin is modified by inputting body measurements or a 3-D body scan of an indi-
vidual is imported into the system, (b) mechanical properties for a variety of fabrics are stored
in the system, (c) seams of patterns are paired to identify how they are stitched together, and
(d) the patterns are placed and curved around the virtual mannequins where virtual stitching
and draping are then performed.

3-D Virtual Fit Technology Use in the Apparel Industry


A few retailers who develop apparel products are successfully piloting the use of 3-D virtual fitting
technology for their prototype development stage (Salmon, 2014; Yoon, 2013). Target and Kohl’s
are utilizing the 3-D Runway Creator for PDS of Optitex as a tool to communicate with their off-
shore vendors (Yoon, 2013). Elaine Hamblin, Manager of Technical Development in the Global
Innovation Research and Development Department at Levi’s, is currently piloting the use of 3-D
virtual technology for designing (Salmon, 2014).
According to a report by Salmon (2014), 3-D simulation technology has reduced the timeline of
the apparel design process by as much as two thirds because this technology allows apparel manu-
facturers to produce fewer physical samples for each garment before production, down now to two
samples from a previous average of three to five. When sample production is conducted offshore, the
benefits of 3-D simulation technology, ‘‘speed,’’ and ‘‘overall lower cost’’ are ever greater because
the number of real prototypes that need to be sent back and forth decreases.
3-D virtual technology helps improve both design and fit communication between designers, pat-
ternmakers, and technical designers (Salmon, 2014; Yoon, 2013). They can all look at the same ini-
tial samples at the same time and edit them in one place, thus giving them the ability to drastically
reduce the time they spend developing their final products and their costs.
In addition to the production field, 3-D virtual fitting technology is a tool apparel product devel-
opers can use to communicate with their customers on the best design and fit. Lori Coulter Swim-
wear provides this service to show each customer examples of virtual custom-fitted swimwear on her
3-D body scan (Lori Coulter, 2013). Customers can communicate their preferences to the company
before their custom clothing is produced.
Online apparel product developers have also utilized 3-D virtual fitting technology to visualize
clothing fit for their customers (i-Fashion Center for Apparel Technology, 2007; My Virtual Model
Inc., 2012; Optitex, 2014). Online sales in clothing are growing at a dramatic pace, accounting for
USD 23.2B in 2010 (U.S. Census Bureau, 2012). However, the online apparel market is still smaller
than other product categories that are sold online. One of the most frustrating aspects of online shop-
ping for clothing is determining the precise fit of the clothing. To address this issue, My Virtual
Model Inc. developed their product, which they describe as software service technology. The tech-
nology allows customers to create personalized 3-D virtual avatars and put on virtual outfits being
sold in the market (My Virtual Model Inc., 2012). My Virtual Model Inc. has licensed their technol-
ogy to apparel companies, such as Lands’ End, Sears, H & M, and Speedo USA. On the My Virtual
Model Inc. website, customers can mix and match 798 tops, 2,653 outerwear garments, 170 shoe
styles, and 111 underwear styles from the apparel companies to whom My Virtual Model Inc. has
licensed the technology (Do, 2010).
In the online mall launched by i-Fashion Technology Center (2007), customers can create
their own parametric virtual avatars by inputting self-reported measurements. Customers can
also visit an office where a 3-D body scanner is installed and obtain their personal 3-D body
scans. They can then automatically import these 3-D body scans to become their personal 3-D
virtual models. Although the My Virtual Model Inc. website focuses on the visualization of

Downloaded from ctr.sagepub.com by guest on September 8, 2015


Song and Ashdown 317

clothing, the i-Fashion mall delivers fitting reports for each customer, including ease amounts
and degrees of pressure.

3-D Virtual Fit Research Studies


Several researchers have analyzed the effectiveness of the current level of technology for visualizing
garment fit accurately, even though most research to date has focused on the virtual and the real fit of
skirts. Kang and Lee (2010) compared the real and the virtual fits of shirred skirts by fabric type
(wool and polyester satin) and shirring amount (1.5, 2, and 2.5 times waist girth). They found that
as the shirring amount increased, the virtual skirt dropped more than the real skirt did. The visuali-
zation of a skirt made from polyester satin was less effective than the visualization of a woolen skirt.
Koo and Suh (2009) compared the fit of real flared skirts to that of virtual flared skirts by grain line
directions (lengthwise, crosswise, and bias) and polyester weight (light, medium, and heavy).
Medium- and heavy-weight flared skirts showed more differences in frequencies of billows and their
locations than did the lightweight skirts. Flared skirts cut in a bias direction showed the fewest dif-
ferences between virtual and real skirts when compared to skirts cut on a lengthwise or a crosswise
direction.
In terms of virtual pants fit, few studies have been undertaken to verify the virtual fit infor-
mation for pants (Kim & LaBat, 2013; Lee, Nam, Cui, & Choi, 2007), and survey results also
have indicated that pants are the most challenging clothing item for a good fit (Eccles, 2011).
Lee et al. (2007) classified body types into three categories (lean, normal, and obese) based on
the Body Mass Index (BMI) readings. They recruited a representative subject for each body
type, developed a pair of pants for lean type, a pair of pants for normal type, and a pair of
pants for obese type. They put all three-pant types on each participant’s parametric avatar
to create three different fit statuses (tight, adequate, and loose). They found that virtual gar-
ments with adequate size represented a silhouette and fit more accurately than garments with-
out size did. It was also determined that the 3-D virtual fit varied depending on three fit
statuses (tight, adequate, and loose); however, the effect of the body type variable on the vir-
tual fit was not seen.
Kim and LaBat (2013) examined the customers’ evaluations of the accuracy of 3-D simulation
technology used in virtual, online shopping scenarios. Thirty-seven female participants were
recruited, and one of the pants (Misses sizes 2–20) was virtually placed on each participant’s 3-D
scan. When comparing the virtual fit evaluation to the real fit evaluation at 13 locations, the parti-
cipants tended to evaluate the virtual fit of the abdomen, back thigh, and front crotch more favorably
than when they were evaluating the actual pants.

Research Objectives and Research Justification


The researchers in the current study investigated whether 3-D virtual fitting technology can ade-
quately/accurately visualize the fit and silhouette of a pants clothing item by comparing real to vir-
tual pants. This study had three specific objectives:

1. Identify the accuracy of 3-D virtual fitting technology for a pants item:
a. analyze the similarities between real and virtual fit depending on fit locations, lower
body shapes, and fit status (overall analyses) and
b. assess the similarities, if any, between real and virtual fit, using lower body shapes and
fit status at each designated fit location (specific analyses).
2. Investigate issues regarding virtual fabric expression and use of 3-D body scan avatar for fit
assessments.

Downloaded from ctr.sagepub.com by guest on September 8, 2015


318 Clothing and Textiles Research Journal 33(4)

We considered that 3-D virtual fit can be visualized differently depending on three factors,
namely, (a) fit status, (b) lower body shapes, and (c) fit locations. We considered fit status (e.g.,
tight/good/loose, or short/good/long) as an influential variable on the final visualization of the fit
and silhouette of the virtual pants. Among the findings in past studies, Lee et al. (2007) found that
garments with adequate size (good fit) represented a more similar fit than garments without size
(tight or loose fit). However, that study also had its limitations, as it relied only on the fit data for
three participants (lean, normal, and obese type) and three statuses for different fit (tight, good,
or loose). In addition, the researchers did not analyze fit at each location, but only evaluated a gen-
eral degree of similarity between virtual and real fit using a 5-point scale (1 ¼ strongly dissimilar to
5 ¼ strongly similar). Therefore, it was not fully possible to analyze how certain fit statuses affected
the most accurate representation of virtual fit at each fit location.
Few researchers have assessed the accuracy of 3-D virtual fitting software in terms of fit analysis
from the viewpoint of patternmakers or technical designers, especially for the visualization of pants.
This study should provide valuable new information for software developers and apparel profession-
als who plan to use the technology.

Conceptual Framework of the Body Shape Analysis Method


The researchers in the current study compare real fit with virtual fit based on lower body shapes
(curvy, hip tilt, and straight shapes). As a conceptual framework for classifying lower body shapes,
the researchers utilized Song and Ashdown’s (2011) body shape categorization. Lower body
shapes, in past studies, were limited to classifying the degree of buttocks prominence into two or
three groups, using visual judgment or a simple calculation of waist-to-hip proportions (Petrova
& Ashdown, 2008; Schofield, Ashdown, Hethorn, LaBat, & Salusso, 2006). Song and Ashdown’s
method (2011) used a multidimensional body shape analysis applying a large anthropometric data
set (SizeUSA) and precise statistical methods (principal component analysis and cluster analysis)
to classify shapes based on a comparison of multiple body dimensions. Based on 16 body measure-
ments, the principal component analysis allowed for extracting five components that were then
sorted into three groups using cluster analysis (see Figure 1a). These five components were PC1
(drop measurements related to waist-to-top-hip silhouette), PC2 (drop measurements related to
top-hip-to-hip silhouette), PC3 (drop measurements related to buttocks prominence), z-score 1
(z-score for the drop between front abdomen depth and waist front depth—a measurement related
to abdomen prominence), and z-score 2 (z-score for the drop between front abdomen depth and front
hip depth—a measurement related to slope from abdomen point to front hip point).
Using cluster analysis, Song and Ashdown (2011) sorted the population into three body shape
groups. Figure 1b shows the means of each component of these three groups. Figure 1c presents the
front silhouettes and profile sketches for each group. The main body shape characteristics of the
three shape groups are thus the following:

 Group 1 (curvy shape)—The curviest silhouette between waist level and hip level, and the
most prominent abdomen silhouette;
 Group 2 (hip tilt shape)—The most prominent buttocks and a lower body tilted toward the
back; and
 Group 3 (straight shape)—A noncurvy silhouette and less prominent buttocks.

Method
The data collection for the current study included four major stages: (a) development of and con-
struction of custom pants, (b) collecting photographs of participants in pants, (c) development of

Downloaded from ctr.sagepub.com by guest on September 8, 2015


Song and Ashdown 319

Figure 1. (a) Five variables identified from the principal component analysis, (b) graph showing the means of
the PC and z-score variables used for cluster analysis across three body shape groups, and (c) front and side
silhouettes of three body shape groups.

3-D virtual try-on files from pants patterns, and (d) visual analysis of the real and virtual pants fit by
expert judges.

Recruiting of Fit Models


Female participants aged 18–35 with a hip girth of less than 123.4 cm (48.6 in; American Society for
Testing and Materials [ASTM] Missy size 20) were recruited in Ithaca, NY, in January 2010 by fliers
distributed throughout the university campus. We scanned participants twice in close-fitting tank
tops/leggings, using a VITUS/XXL 3-D Body Scanner by Human Solutions. From these two scans,
we chose the scan that delivered better quality. Scan files were converted to the rbd format, and 44 body
measurements to classify the body shapes and drafting pants were derived using the automatic measur-
ing software NX16 (Version 6) from [TC]2. Because this measuring software was used for the SizeUSA
survey, the same landmarks and measuring locations found in the SizeUSA data could be detected. They
became the base data for Song and Ashdown’s 2011 study. Crotch height was measured manually only
since the 3-D scanner could not reliably measure crotch areas due to its technical limitations.
The researchers in the current study used Song and Ashdown’s (2011) body shape categorization
method to classify lower body shapes (curvy, hip tilt, and straight shape). Principal component anal-
ysis and cluster analysis (Song & Ashdown, 2011) were both run to identify the body shape group for
each participant. Participants were recruited across the full range of size and shape groups. Of the 81
participants recruited, 61 (Group 1, curvy shape ¼ 21; Group 2, hip tilt shape ¼ 21; and Group 3,
straight shape ¼ 19 participants) were identified for the current analysis.

Downloaded from ctr.sagepub.com by guest on September 8, 2015


320 Clothing and Textiles Research Journal 33(4)

Instrument: Development of Custom-Fitted Pants


The instrument for this study was a pair of custom-fitted pants. The pants style had two front darts,
two back darts, and a straight silhouette from the abdomen and buttocks to the hem. The pants were
constructed of a stable, medium-weight, 100% cotton twill fabric in a neutral color. For the base
pattern, we chose an industry pattern from a major direct marketing (catalog and Internet) retailer
with a reputation for good fit. Then, a style with fabric properties similar to the fabric used for the
study was chosen. Because of retailer regulations, these patterns could not be officially provided in
their precise detail for the current study. Therefore, we purchased a size 10 (hip size: 101.6 cm, 40
in) of the style, and then the pants were rubbed off. Using Pattern Design 8.3 from Gerber Scientific,
we digitized patterns and graded the base pattern (size 10) from size 2–20 using the same method
described in Grading Techniques for Modern Design (Price & Zamkoff, 1996).
To develop custom pants for each participant, we used the automated custom patternmaking CAD
software AccuMark MTM from Gerber Scientific. For the pants patterns, we identified and num-
bered alteration points that corresponded to five locations judged as necessary for good fit (waist
girth, hip girth, thigh girth, waist-to-crotch length, and inseam length). We manually input each par-
ticipant’s five measurements into the system. From the body charts that we stored, hip girth was
selected as the primary measurement for the selection of each participant’s base size. When gener-
ating custom-made patterns, the system selected the size of the base pattern from the graded nest for
each individual by comparing her hip girth to the range of hip girths for all the body sizes on the body
chart. After selecting the base size patterns from the graded nests, the system compared the rest of
the measurements with those in the appropriate body charts and calculated the alteration amounts at
each location. After making alterations at waist girth, hip girth, thigh girth, waist-to-crotch length,
and inseam length, the system generated the final custom pants patterns. Based on previous studies
(Apeagyei & Otieno, 2007; Ashdown & Dunne, 2006; Song & Ashdown, 2012), we expected that
the software would give us pants with various fit statuses.

Collecting Photographs of Participants in Pants


The 61 participants (Group 1 ¼ 21, Group 2 ¼ 21, and Group 3 ¼ 19 participants) tried on their
custom pants. We took photographs of them in front, side, and back views in standard posture for
a visual fit evaluation by our expert judges.

3-D Virtual Try-On


We used Optitex PDS 3-D Runway Creator software because many global apparel companies (e.g.,
Nike, Target, and Kohl’s) and different fashion schools (e.g., Parsons School of Design, London
College of Fashion, and Cornell University) use it. We converted the participants’ pants patterns,
generated from the AccuMark MTM system, to a dxf format, and then imported the dxf files to the
Optitex software. No compatibility issues were found.
In 3-D Runway Creator, 3-D virtual models can be created using two methods: (a) importing the
3-D body scan (obj file) or (b) creating a parametric model by inputting body measurements. Lim
and Istook (2011) found that the parametric model was different from the actual body shape. From
the side view, the waist, abdomen, and hip area of the parametric model had more dimensions than
the actual body. Therefore, we chose method (a) or the 3-D body scan. Using NX16 software (Ver-
sion 6), we converted the 3-D scan files to an obj format and imported them into the 3-D Runway
Creator program. We then inputted the fabric properties that were tested using the Kawabata fabric
evaluation system to the same software. We virtually tried each participant’s pants on her 3-D body
scan (obj format) and screen-captured the front, side, and back views.

Downloaded from ctr.sagepub.com by guest on September 8, 2015


Song and Ashdown 321

Visual Analysis of Real Versus Virtual Pants and Their Fits


For the visual fit evaluation of the pants and an assessment of the similarities between real and vir-
tual fit, we recruited nine expert fit judges who have worked as designers or technical designers in
the women’s apparel industry for 8 years or more.
We developed a rating instrument for the study that contained two sections. First, we provided
front, side, and back view photos of a body in the pants and asked each judge to rate that fit at 17 loca-
tions, such as waist ease (front and back), abdomen ease, hip ease, crotch ease (front and back), thigh
ease (front and back), hem width—distance between side seams, hem depth—distance between left
side seam to front and back (front and back), crease line at hem, waist placement (front and back),
crotch placement, side seam placement, and pants length. A 3-point scale was used to rate this fit. The
end points of each scale varied depending on the actual fit locations. For example, tight (1)–loose (3) or
short (1)–long (3). The middle value of this scale, or a rating of two, represented a good fit.
Each expert judge was also asked to rate the number of stress folds caused by misfit at four loca-
tions: between the waist and the crotch for the front and the back and between the crotch and the
knee for the front and the back. The number of stress folds were indicated using a 3-point scale rang-
ing from 1 (no folds) to 3 (many folds) for both front and back.
Second, the experts compared the real fit (based on a photo of a body in pants) to the virtual
fit (based on the screen-captured image of a scanned body in virtual pants). They then evalu-
ated the degree of similarity between the real and the virtual fit on a 5-point Likert-type scale,
ranging from 1 (very different) to 5 (very similar) at each designated location. If the real fit and
the virtual fit were different, the judges reported through written feedback how these two fits
were different.
After the judges completed the 61 questionnaires, we interviewed them for (1) how the fabric sur-
face looked different from that of the real pants and (2) any issues regarding the 3-D body scan that
was used as an avatar for fit assessment.

Analysis Method for Comparing the Fit of Virtual and Real Pants
Objective 1A (overall analysis). The purpose of Objective 1A was to analyze for any similarity between
real and virtual fit for three body shapes, two fit statuses (good or poor fit), and 20 fit locations. Orig-
inally, a 3-point scale was used to rate the fit (e.g., tight [1], good [2], and loose [3]). However, to
judge whether the similarity actually varied based on good or bad fit, the researchers recoded the
scale choice for good fit (2) as good fit (0) and the rest of the scale choices (e.g., tight [1] and loose
[3]) as bad fit (1).
Then, a multiple regression model was used to calculate heteroskedasticity-consistent standard
errors and co-variance by using the Eview statistics package. A dependent variable was found to be
‘‘similarity between real and virtual fit’’ (a 5-point scale). Three independent variables (fit status, body
shape, and fit location) were changed to dummy variables. There were several reasons why the mul-
tiple regression analysis, using dummy variables, was used instead of analysis of variance (ANOVA),
and these could have also been considered an acceptable analysis method for this study. First, the use
of dummy variables as predictors in regression provides a simple demonstration of the fundamental
equivalence between ANOVA and multiple regression analysis. An advantage of the multiple regres-
sion analysis is that it allows the use of more than two variables as predictors and does so simultane-
ously (Warner, 2008). Second, the regression analysis can include the interaction between two
dummy-coded variables to allow for a potentially clearer and more precise interpretation of results.
The process of creating dummy variables is the following: When there are multiple groups (num-
ber of groups ¼ k), group membership can be represented by scores from a set of (k  1) dummy
variables. The independent variables in the regression model were:

Downloaded from ctr.sagepub.com by guest on September 8, 2015


322 Clothing and Textiles Research Journal 33(4)

 nineteen dummy-coded ‘‘fit location’’ variables (‘‘front waist placement’’ was coded as 0 on
all 21 dummy variables);
 two dummy-coded ‘‘body shape’’ variables (‘‘body shape 1’’ was coded as 0 on the two
dummy variables);
 a dummy-coded ‘‘degree of fit’’ variable (‘‘good fit’’ was coded as 0 and ‘‘bad fit’’ as 1);
 the interaction between ‘‘body shape 2’’ and ‘‘the degree of fit’’ variables; and
 the interaction between ‘‘body shape 3’’ and ‘‘the degree of fit’’ variables.

Objective 1B (specific analysis). The purpose of Objective 1B was to assess the similarity between real
and virtual fit based on three lower body shapes and three fit statuses (e.g., tight, good, or loose) at
each fit location. We conducted a multiple regression model for Objective 1B, but two differences
from the model for Objective 1A remained. First, we conducted a multiple regression model at each
fit location. Second, we judged whether the similarity between real and virtual fit varied depending
on a more detailed fit status (e.g., tight, good, or loose). We thus kept a 3-point scale. The indepen-
dent variables at each fit location were:

 two dummy-coded ‘‘body shape’’ variables (‘‘body shape 1’’ was coded as 0 on the two
dummy variables);
 two dummy-coded ‘‘degree of fit’’ variables (‘‘good fit’’ was coded as 0 on the two dummy
variables);
 the interaction between the variables ‘‘body shape 2’’ and ‘‘the degree of fit’’; and
 the interaction between the variables ‘‘body shape 3’’ and ‘‘the degree of fit’’ variables.

Results
Overall Validity of the Virtual Fitting Software
Table 1 shows the results for the multiple regression analysis. The regression indicated that the pre-
dictors explained 10.04% of the variance, R2 ¼ .10, F(23, 5,430) ¼ 22.46, p < .01. The coefficient
(b) for each of the dummy variables in the regression model, using multiple dummy variables, rep-
resents the difference between the mean of the corresponding group and the mean of the comparison
group, whose code was 0 for all the dummy-coded variables. Therefore, the results can be inter-
preted as follows:

1. Fit status (good fit was coded as 0 and bad fit as 1). The virtual images of pants with a good
fit represented a silhouette and fit more effectively than did pants with a bad fit (b ¼ .35,
t ¼ 6.52, p < .01).
2. Body shape group (‘‘body shape 1’’ was coded as 0 on all the dummy variables). The coeffi-
cient score (b ¼ .16) for Group 2 corresponds to the mean difference between Groups 2 and
1, the comparison group whose codes were all 0 (t ¼ 3.15, p < .01). The mean scores for
Group 3 were not significantly different from those for Group 1. Therefore, the virtual repre-
sentation of fit for the pants from Group 2 was the least effective of any of the analyzed groups.
3. Fit location (front waist placement was coded as 0 on all the dummy variables). The coefficient
score (b) for each fit location corresponds to the difference between the mean for the similarity
score for each fit location and the mean score for front waist placement, the comparison fit
location whose codes were all 0. Comparing these coefficient scores enabled an identification
of the fit locations that were best represented by the 3-D virtual program. In Table 1, four fit
locations (back waist placement, crotch placement, hem width, and stress folds on the back of

Downloaded from ctr.sagepub.com by guest on September 8, 2015


Song and Ashdown 323

Table 1. Results of Regression Analysis for Overall Assessment of Validity of 3-D Virtual Fitting Software.

Variable b SE t p

Fit status Good Fit Dummy* .35 .05 6.52 .00


Body shape Group 2 Dummy* .16 .05 3.15 .00
group Group 3 Dummy .08 .05 1.54 .12
Interaction Interaction (Fit Dummy  Group 2 Dummy) .06 .08 0.75 .45
Interaction (Fit Dummy  Group 3 Dummy) .03 .08 0.44 .66
Fit location Waist placement (B) .08 .11 0.69 .49
Waist ease (F)* .85 .11 7.62 .00
Waist ease (B)* .70 .11 6.30 .00
Abdomen ease* .58 .11 5.32 .00
Hip ease* .72 .11 6.53 .00
Crotch placement .13 .11 1.16 .25
Crotch ease (F)* .59 .11 5.20 .00
Crotch ease (B)* .71 .11 6.37 .00
Thigh ease (F)* .71 .11 6.47 .00
Thigh ease (B)* .68 .11 6.11 .00
Pants length* .27 .11 2.50 .01
Side seam placement* .83 .11 7.53 .00
Hem width—distance between side seams .03 .11 0.30 .76
Hem depth (F)—distance between left side seam to .35 .11 3.12 .00
front crease line at hem
Hem depth (B)—distance between left side seam to .59 .11 5.33 .00
back crease line at hem*
Stress fold amount between waist and crotch (F)* .23 .10 2.26 .01
Stress fold amount between crotch and knee (F)* .33 .11 2.94 .00
Stress fold amount between waist and crotch (B) .19 .11 1.78 .07
Stress fold amount between crotch and knee (B) .06 .10 0.60 .55
Constant 2.40 .09 28.05 .00

R2 .10 Mean dependent 2.84


variable
Adjusted R2 .10 SD dependent 1.22
variable
SE of regression 1.16 Akaike info 3.14
criterion
Sum squared 7,307.71 Schwarz criterion 3.17
residual
Log likelihood 8,543.15 F value 22.46
Durbin-Watson 1.20 p value .00
stat
Note. SE ¼ standard error. (F) ¼ front, (B) ¼ back; SD ¼ standard deviation.
*p < .01.

the pants) did not have significantly different scores from the scores for front waist placement
(p < .01). However, the results for the rest of the locations were significantly different.

A comparison of these coefficients is shown in the graph in Figure 2. The coefficient score (b) for
front waist placement was 0, the mean score for similarity was 2.4. Therefore, the mean score for
each location was b þ 2.4. Locations where the 3-D software did not represent fit well (mean score
of similarity < 3) were waist placement, crotch placement, hem width, stress folds on the back of the

Downloaded from ctr.sagepub.com by guest on September 8, 2015


324 Clothing and Textiles Research Journal 33(4)

Figure 2. Coefficient scores of fit locations.

pants, stress folds between the crotch and knee, stress folds between the front waist and front crotch,
and pant length.

Validity of 3-D Virtual Fitting Software at Each Fit Location


At each fit location, a multiple regression analysis was used to assess the similarity between real and
virtual fit, depending on lower body shapes and fit status. The results from the multiple regression
analysis are shown in Table 2. We noted b values at cells where they were significant (p < .01). The
difference between virtual fit and real fit is also shown in Table 2.
At the waist, abdomen, front crotch, and thigh, the pants with appropriate ease were represented
more accurately than the pants with a tight or loose fit (p < .01). However, at hip (b ¼ .78, p < .05)
and back crotch ease (b ¼ .46, p < .05), pants with either good or loose fit were represented more
accurately than were pants with tight fit (p < .01). It was found that 57.6–76.9% of the cases in which
expert judges considered virtual and real fit to be different were cases in which the virtual pants
appeared looser than the real pants.
In terms of the number of stress folds, fewer stress folds generally meant an increase in the simi-
larity between real and virtual fit. It was found that 6.0–28.1% of the cases where expert judges con-
sidered virtual and real fit to be different were cases where the virtual pants showed no stress folds
even though those pants did have small stress folds. In 65.6–77.7% of these cases, the virtual pants
showed fewer stress folds than the real pants did.
Regarding hem width, the similarity between the two designated images, namely, virtual pants
and real pants, generally increased as the hem became wider. When hem width was seen from the
front, 95.8% of the cases where expert judges considered virtual and real fit to be different were
cases where the virtual pants had wider hems than the real pants did. From the side, however,
73% of the cases where expert judges considered virtual and real fit to be different were cases where
the virtual pants had narrower hems than the real pants.

Downloaded from ctr.sagepub.com by guest on September 8, 2015


Table 2. Assessment of Validity of 3-D Virtual Fitting Software at Each Fit Location.

Regression analysis Frequency (%)

Body shape (b) Fit status (b) Virtual pants were __ than real one
2
Variable R F (8, 537) Constant (b) 2 3 Tight Loose Tighter Looser

Ease Waist ease (F) .117 8.98** 3.54 .87** .94* 17.4 82.8
Waist ease (B) .267 24.54** 3.71 .98** 1.28** 33.3 66.8
Abdomen ease .083 6.06** 3.49 .37 * .48* .95** .68** 31.0 69.0
Hip ease .108 8.11** 3.58 .78** 21.0 79.1
Crotch ease (F) .092 6.87** 3.42 .49** .77** .79** 37.7 62.5
Crotch ease (B) .203 2.88** 3.36 .46* 35.9 64.1
Thigh ease (F) .088 6.50** 3.55 .44** .69** .74** 31.9 68.1
Thigh ease (B) .035 2.45* 3.43 .81* .44* 33.3 66.7
Virtual pants had __ wrinkles than real one
Little Much No Less More
Stress fold amount Between waist and crotch (F) .132 10.23** 2.97 .72** 28.1 65.6 6.4
Between waist and crotch (B) .077 5.62** 2.99 .73** 23.7 68.4 8.1
Between crotch and knee (F) .093 6.93** 3.22 .31** .32** .61* .53* 6.0 77.7 16.3
Between crotch and knee (B) .167 13.44** 2.96 .77** .70** 23.5 66.4 10.2
Virtual pants had ___ hem than real fit
Narrow Wide Narrower Wider
Hem width/depth Hem width .078 5.73** 2.53 .61** 4.2 95.8
Hem depth (F) .037 2.58*** 3.08 73.0 27.1

Downloaded from ctr.sagepub.com by guest on September 8, 2015


Hem depth (B) .899 598.00*** 3.00 1.24** 1.18** 73.1 27.0
Virtual pants had ___ placement than real fit
Low High Lower Higher
Placement/length Waist placement (F) .055 3.89** 2.51 81.3 18.8
Waist placement (B) .038 2.65* 2.77 .48* .62* 74.1 26.0
Crotch placement .065 4.55** 3.05 .52** 0.65** 41.4 58.7
Virtual pants had ___ length than real fit
Short Long Shorter Longer
Pants Length .108 8.15*** 3.05 .72* 1.36** 30.4 69.7
Note. Coefficients (b) noted at cells they were significant; (F) ¼ front; (B) ¼ back.
*p < .01. **p < .05. ***p < .001.

325
326 Clothing and Textiles Research Journal 33(4)

For front waist placement, the results from this analysis were not significantly different, depend-
ing on the individual fit status. The constant (b) was 2.51, which was the lowest score among the
different fit locations. To sum up, front waist placement of the virtual pants appeared to be different
from the real pants, regardless of fit status. For back waist placement and crotch placement, the pants
with ‘‘appropriate and high placement’’ were represented more effectively than were the pants with
‘‘low placement.’’ However, for pants with an ‘‘appropriate length,’’ the virtual image represented
the silhouette more effectively than did the pants with ‘‘short or long length.’’ Virtual pants also gen-
erally had a lower waist placement (74.1–81.3%). Accordingly, the virtual pants appeared to have a
longer in-seam length than that of the actual pants (74.1–81.3%). For side seam placement, the final
results were not significantly different but still were dependent on fit status. The side seam place-
ment of virtual pants generally appeared more forward than that of the actual pants (73.0%).
The virtual pants for the participants in Group 2 represented the greatest divergence from the fit
of their actual pants among all the body shape groups. Group 2 had a lower body tilted toward the
back (see Figure 1). Accordingly, between the abdomen and front knee level, Group 2 pants had
more gaps between the body and the pants; therefore, the pants also had more stress folds than those
of the other groups.

Fabric Expression and 3-D Body Scan


With respect to fabric, the judges indicated three issues. First, they noted that the 3-D virtual soft-
ware could only express stress folds caused by a very tight or a loose misfit. It could not express
small folds caused by a slight misfit (Figure 3a).
Second, the judges indicated that fabric color and texture were different from actual fabric. They
found that 3-D virtual fabric color and texture looked shinier, smoother, and thicker than actual fab-
ric. In this instance, they looked at a stable medium-weight cotton twill fabric in a neutral color.
Originally, we scanned the fabric and attempted to apply the scanned surface texture to the virtual
pant models. However, the stress folds were not visible when we applied the scanned texture, so we
then selected a color similar to the actual fabric from the available color palette.
Third, the judges indicated that the direction of stress folds on the 3-D virtual pants was different
from that on the real pants, especially between the crotch and the hem. The reason for this difference
may have been that the hem from the front view was wider, whereas the hem width from the side
view was narrower than that of the real pants (Figure 3b). Because the spread profile for the bottom
edge of the pants was different for the real and the virtual pants, the direction of the wrinkles was
also affected.
In terms of the issues found for the 3-D body scan as an avatar, the fit of the virtual pants was also
influenced by the posture of the 3-D scanned model. Even when the actual fit was good, the virtual
fit did not look precise if the 3-D virtual model was displayed in a slightly unbalanced posture.

Conclusion, Discussion, and Future Research Opportunities


Although 3-D technologies have been adopted by only large companies such as Target and Levi’s
and only on a relatively small scale, these companies are successfully piloting the use of 3-D simu-
lation technology for designing and prototyping clothing samples (Salmon, 2014; Yoon, 2013).
Other retailers are beginning to understand the implications of this 3-D simulation technology, and
thus broad adoption of the concept will be inevitable as time passes, given how impactful the ben-
efits of ‘‘speed’’ and ‘‘lower overall production costs’’ are for the entire industry (Salmon, 2014).
We investigated whether 3-D virtual fitting technology can effectively visualize the fit and sil-
houette of pants in more general terms by using three factors, namely, fit status, lower body shapes,
and fit locations from both designers and technical designer perspectives. Of the past research

Downloaded from ctr.sagepub.com by guest on September 8, 2015


Song and Ashdown 327

Figure 3. Issues on virtual fit expression (a) the software expressed less wrinkles than the actual amount,
(b) the hem width from the front view was wider than that of real pants, but the hem width from the side view
was narrower than that of real pants.

studies, only a few researchers investigated the accuracy of 3-D garment simulation technology for
visualizing garment fit, especially for pants. Lee et al. (2007) simply analyzed the degree of simi-
larity between real and virtual fit, using but three participants (lean, normal, and obese body type)
and three overall different pant fit categories (tight, adequate, and loose). However, we developed 61
individual pant samples for three lower body types using an automated custom fit program that
allowed us to obtain pants with various fit problems. We also asked the fit judges to analyze the pant
fit so we could analyze how 3-D representation was different based on a variety of fit statuses.
Kim and LaBat (2013) also evaluated the accuracy of 3-D simulation technology. They selected
36 general consumers as evaluators, so the analysis was from the customer perspective in the online
shopping scenario. However, we asked expert fit judges who actually worked as designers or tech-
nical designers to evaluate the similarities between the real and the virtual fit. Another difference

Downloaded from ctr.sagepub.com by guest on September 8, 2015


328 Clothing and Textiles Research Journal 33(4)

between Kim and LaBat’s study and the current one is that Kim and LaBat’s used a 7-point Likert-
type scale with end points being extremely poor fit (1) and excellent fit (7), while we used a Likert-
type scale with different end points for each scale based on fit locations (e.g., tight [1]-loose [3] or
short [1]-long [3]). The fit judges in the current study additionally evaluated the degree of similarity
between real and virtual fit using a 5-point Likert-type scale, ranging from 1 (very different) to 5
(very similar) at each location. Therefore, we could analyze how 3-D representation was different
depending on a variety of fit statuses at each designated fit location.
From these results, we concluded that the overall accuracy of virtual fitting technology is mod-
erately good for use, especially for pants with good fit, but not to the extent that apparel professionals
could fully use it as a visual fit analysis tool. We noted five main issues and offer suggestions on
ways to improve the 3-D simulation technology.
First, 3-D virtual pants were not donned at the right place, especially at the waist. Specifically,
waist placement of the virtual pants appeared to be different from that of the real pants regardless
of fit status. Most of the cases in which virtual fit and real fit looked different were when the waist
placement of the virtual pants was lower than that of the actual pants. This problem made it difficult
for the judges to judge crotch placement and pant length. Accordingly, it became difficult to judge
ease at the waist, hip, and crotch. In preparing the questionnaire, we also had difficulty in donning
the virtual pants because the pant waist placement changed depending on the initial alignment of the
pant patterns on the body. To solve this problem, further tools should be developed to let apparel
professionals mark the landmarks of waist location on the 3-D avatar before donning the pants and
to let the 3-D virtual program place the waist of the pants on these landmarks accurately.
Second, the 3-D virtual software generally represented less ease than the actual pants, and the
software could not express small stress folds from slight tightness and looseness. It was found that
the 3-D virtual image had no stress folds (6.0–28.1% of the cases in which expert judges considered
virtual fit and real fit to be different), or fewer stress folds than the real pants demonstrated (64.8–
74.3%). At the prototype development stage in the industry, the prototypes may only have slight fit
problems because the company will alter patterns that have already been shown to fit well. The 3-D
virtual program should be improved to express small stress folds from such slight misfits, especially
those between the waist and the crotch.
Third, the 3-D virtual program could not represent the silhouette of the pants, especially from the
crotch to the hem. Toward the hem, the front silhouette of the virtual pants spread wider than that of
the real pants, while the profile view was narrower. Currently, a few apparel companies, especially
Target and Kohl’s, have tested the utilization of 3-D virtual programs to view garment silhouettes
before producing their prototypes. Moreover, online apparel companies have utilized the technology
to visualize clothing silhouettes for their customers to judge before they purchase the clothing
(i-Fashion Center for Apparel Technology, 2007; My Virtual Model Inc., 2012). We found, how-
ever, that this technology may not yet be developed enough to represent style and silhouette appro-
priately for customers to make their final purchase decisions more easily.
Fourth, the unique finding of this study was that virtual pant fit is affected by body shape. The
results show that the virtual representation of pant fit for Group 2 (hip tilt and the most prominent
buttocks shape) was the least effective of any of the body shape groups, especially at front crotch and
thigh ease and for back waist placement. This finding means that if the samples are virtually fitted on
the fit model avatar with this shape, then virtual fit is not represented well. We determined that hip
tilt and the most prominent buttocks shape cause the loose fit and wrinkles at the front crotch and
front thigh, and a gap between the body and the pants at the back waist due to this loose fit.
One of the most critical issues to be addressed for the virtual program is the reality/clarity of fab-
ric color and texture. If the scanned fabric image is filled into the pants, then stress folds are
obscured. However, when the color is filled from the pallet, the 3-D virtual fabric looks smoother
and thicker than the actual fabric. Apparel professionals want to see virtual clothing filled with

Downloaded from ctr.sagepub.com by guest on September 8, 2015


Song and Ashdown 329

an actual fabric image to test the harmony between fabric and design. Thus, the software should be
improved to show stress folds if the scanned image is used to apply texture and color to the virtual
garments.
However, this study is somewhat limited, as the results are based on only one kind of fabric and
one pant style. Further, the study was based on a photo of a body in pants (for real fit) and a screen-
captured image of a scanned body in virtual pants (for virtual fit). In future studies, as fit judges are
trained in the use of 3-D simulation technology, they can then analyze virtual fit by zooming in on
and rotating the 3-D image. They can also participate in fit sessions with a live fit model in clothes.
The results of such a comparison between virtual and real fit will then be more accurate, reliable, and
thus more accepted.
Early adopters such as Target and Kohl’s already understand the weak points of this new tech-
nology, but they keep using it because they know how impactful the benefits of speed and lower
overall production costs can be for their companies. Target plans to use virtual technology to
develop 80–90% of all their apparel products by 2018. If more researchers in the apparel product
development field pay closer attention to the value of 3-D virtual fit representation and provide
appropriate feedback to 3-D virtual fitting software developers, this new technology can be devel-
oped further and address the limitations. Then more apparel companies will be willing to use 3-D
virtual fitting technology as a valuable tool, producing tremendous savings in company production
time and costs, as well as ongoing benefits to their customers.

Declaration of Conflicting Interests


The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publi-
cation of this article.

Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.

References
Apeagyei, P. R., & Otieno, R. (2007). Usability of pattern customizing technology. Journal of Fashion Mar-
keting and Management, 11, 349–365.
Ashdown, S. P., & Dunne, L. (2006). A study of automated custom fit: Readiness of the technology for the
apparel industry. Clothing and Textiles Research Journal, 24, 121–136.
Carignan, M., Yang, Y., Magnenat-Thalmann, N., & Thalmann, C. (1992). Dressing animated synthetic actors
with complex deformable clothes. Computer Graphics, 26, 99–104.
Do, W. H. (2010). The comparison of user preference on domestic versus a foreign 3D virtual try-on system.
Journal of the Korean Society of Clothing and Textiles, 34, 1184–1196.
Eccles, L. (2011, January 22). Why most women can only dream of perfect jeans: Two-thirds can’t find a pair
that fit. The Daily Mail. Retrieved from http://www.dailymail.co.uk/news/article-1349501/Why-women-
dream-perfect-jeans-Two-thirds-pair-fit.html
i-Fashion Center for Apparel Technology. (2007). i-Fashion mall. Retrieved from http://www.ifashion.or.kr
Kang, I., & Lee, S. (2010). A study on the shape of shirring using the 3D virtual clothing system. The Journal of
Korean Society of Clothing and Textiles, 34, 1111–1125.
Kim, D. E., & LaBat, K. (2013). An exploratory study of users’ evaluations of the accuracy and fidelity of a
three-dimensional garment simulation. Textile Research Journal, 83, 171–184.
Koo, M. R., & Suh, M. A. (2009). A study on the shape of hem line of semi-flare skirts according to a cutting
angle - Based on a comparison between real clothing and 3D virtual clothing. The Research Journal of the
Costume Culture, 17, 499–511.
Lafleur, B., Magnenat-Thalmann, N., & Thalmann, D. (1991). Cloth animation with self-collision detection. In
T. Kunii (Ed.), Modeling in computer graphics (pp. 179–187). Tokyo: Springer Japan.

Downloaded from ctr.sagepub.com by guest on September 8, 2015


330 Clothing and Textiles Research Journal 33(4)

Lee, J., Nam, Y., Cui, M. H., & Choi, K. M. (2007). Fit evaluation of 3D virtual garments. Lecture Note in
Computer Science, 4559, 550–558.
Lee, M., & Sohn, H. (2011). Study on cases of the application of 3D apparel CAD system to domestic and over-
seas fashion education. Journal of the Korean Society of Clothing and Textiles, 35, 1112–1124.
Lim, H. S., & Istook, C. L. (2011). Comparitive assessment of virtual garments using direct and manual avatars.
The Research Journal of the Costumen Culture, 19, 1359–1371.
Lori Coulter. (2013). Our story. Retrieved from http://www.loricoulter.com/pages/our-story
My Virtual Model Inc. (2012). My virtual model. Retrieved from http://corpo.myvirtualmodel.com/index.html
Optitex. (2014). 3D suite. Retrieved from http://www.optitex.com/en/3D-Suite-Create-Garments-Fit-to-Avatar-
Animate
Petrova, A., & Ashdown, S. P. (2008). Three-dimensional body scan data analysis: Body size and shape depen-
dence of ease values for pants’ fit. Clothing and Textiles Research Journal, 26, 227–252.
Price, J., & Zamkoff, B. (1996). Grading techniques for modern design (2nd ed.). New York, NY: Fairchild
Publications.
Salmon, K. (2014). The coming revolution in retail, courtesy of 3D technology. Kurt Salmon. Retrieved from
http://www.kurtsalmon.com/US/vertical-insight/The-Coming-Revolution-in-Retail%2C-Courtesy-of-3D-
Technology?vertical¼Retail&id¼1034&language¼en-us#.U3rGnk1ZpjE
Schofield, N. A., Ashdown, S. P., Hethorn, J., LaBat, K., & Salusso, C. J. (2006). Improving pant fit for women
55 and older through an exploration of two pant shapes. Clothing and Textiles Research Journal, 24,
147–160.
Song, H. K., & Ashdown, S. P. (2010). An exploratory study of the validity of visual fit assessment from three-
dimensional scans. Clothing and Textiles Research Journal, 28, 263–278.
Song, H. K., & Ashdown, S. P. (2011). Categorization of lower body shapes for adult females based on multiple
view analysis. Textile Research Journal, 81, 914–931.
Song, H. K., & Ashdown, S. P. (2012). Development of automated custom-made pants driven by body shape.
Clothing and Textiles Research Journal, 30, 315–329.
Speer, J. K. (2008, November 5). Victoria’s secret: Framing the fit problem. Apparel Magazine. Retrieved from
http://www.apparelmag.com
U.S. Census Bureau. (2012, May 10). E-commerce 2010. U.S. Census Bureau E-Stats. Retrieved from http://
www.census.gov/econ/estats/2010/2010reportfinal.pdf
Volino, P., Cordier, F., & Magnenat-Thalmann, N. (2005). From early virtual garment siulation to interactive
fashion design. Computer-Aided Design, 37, 593–608.
Warner, R. M. (2008). Applied statistics: From bivariate through multivariate techniques. Los Angeles, CA:
Sage.
Yoon, Y. S. (2013, September 17). Representation of real fashion via 3D virtual technology. The Sisa News
Times. Retrieved from http://www.sisanewstime.co.kr/news/articleView.html?idxno¼265

Author Biographies
Hwa Kyung Song earned a PhD degree at Cornell University, and she is an assistant professor in the Depart-
ment of Clothing & Textiles at Kyung Hee University. Her recent research topics include body shape analysis
using 3-D body scan data, comparison of 3-D clothed scan with virtual fitting image, and improvement of auto-
mated made-to-measure patternmaking process.
Susan P. Ashdown is the Helen G. Canoyer professor in the Department of Fiber Science & Apparel Design at
Cornell University. Her research is on the interaction between apparel design and technology, 3-D body scan-
ning, apparel fit and sizing, and changes in apparel design, production, and distribution driven by new
technologies.

Downloaded from ctr.sagepub.com by guest on September 8, 2015

You might also like