MS Word Template for CAD Conference Papers

Computer-Aided Design & Applications, 20(S13), 2023, 111-123

111

Intelligent Clothing Design and Production Integrating CAD and

Virtual Reality Technology

Xianyu Wang

1

and Xiaoguang Sun

2

1

School of Fashion, Henan University of Engineering, Zhengzhou, Henan 451191, China,

[email protected]

2

School of Fashion, Henan University of Engineering, Zhengzhou, Henan 451191, China,

[email protected]

Corresponding author: Xianyu Wang, [email protected]

Abstract. The full name of clothing CAD is computer-aided clothing design, which

refers to the use of electronic computers to assist personnel in clothing design. It

has higher efficiency, better quality, and more accurate positioning than traditional

manual design. In addition, many clothing CAD systems have integrated intelligent

technology, elevating traditional clothing design to the level of intelligent assisted

design. This project initiated the application research of intelligent visual

technology based on DL (deep learning) in ethnic clothing culture and pattern

design. A pattern generative model based on DL is proposed. Our new model uses

a two-layer LSTM (long- and short-term memory) network in the decoding part,

and performs the required mapping through sufficient depth and nonlinear

transformation, instead of placing a simple single hidden layer multilayer

perceptron on the top of the decoder in the original language decoding model.

Research has shown that the algorithm proposed in this paper greatly improves the

pattern matching accuracy of four corner rotation changes, with an accuracy

increase of 56.127%. From the perspective of the highest accuracy of a single

pattern, the traditional algorithm has a maximum accuracy of only 70.066%, while

for some ethnic patterns, the highest accuracy of the improved algorithm in this

article can reach 100%. The language decoding model based on double-layer LSTM

can perform more nonlinear transformations on image and language information,

thereby improving the semantic expression ability of the generated description

statements.

Keywords: Computer Aided Design; Deep learning; Intelligent vision technology;

National costume; Pattern

DOI: https://doi.org/10.14733/cadaps.2023.S13.111-123

1 INTRODUCTION

Clothing CAD technology is a clothing design tool for clothing production and processing

enterprises. It effectively enables clothing engineering technicians to design better and more

Computer-Aided Design & Applications, 20(S13), 2023, 111-123

112

market competitive fashion. Applying clothing design knowledge to the process of clothing design,

achieving optimized and intelligent design while incorporating rich graphic processing functions. It

is an inevitable trend to develop intelligent garment CAD technology for fashion design on the

basis of general garment CAD technology mainly based on geometric modeling. Clothing CAD

design refers to the process of using computer software and hardware to assist in designing

clothing drawings. This technology is very important in clothing production and design, which can

greatly improve efficiency and quality. A clothing CAD system generally consists of two parts:

hardware and software. The hardware includes computers, scanners, printers, etc., while the

software includes clothing design software and CAD drawing software. Abdulahimovna [1]

discussed the clothing CAD system, which can help designers draw various patterns of clothing on

computers.

CAD intelligent clothing refers to the process of using computer software and hardware to

assist in designing clothing drawings. This technology is very important in clothing production and

design, which can greatly improve efficiency and quality. A clothing CAD system generally consists

of two parts: hardware and software. The hardware includes computers, scanners, printers, etc.,

while the software includes clothing design software and CAD drawing software. Clothing CAD

management is a complex process that involves system design, development, implementation,

operation, and maintenance. Only through careful management and practice can the normal

operation of the clothing CAD system be ensured, and the efficiency and quality of clothing design

and production be improved. Clothing CAD is a three-dimensional modeling and display technology

based on virtual reality technology, which can convert planar two-dimensional clothing templates

into three-dimensional clothing samples. This technology is also known as 3D visual stitching

technology. A clothing CAD system generally consists of two parts: hardware and software. The

hardware includes computers, scanners, printers, etc., while the software includes clothing design

software and CAD drawing software. The clothing CAD system can help designers layout and size

various parts of clothing, such as tops, pants, sleeves, etc., in order to make the production

process more efficient. In short, clothing CAD design is an important clothing design technology

that can greatly improve efficiency and quality, helping designers better complete clothing design

and production.

2 RELATED WORK

Du et al. [2] conducted geometric optimization design of the product, effectively solving the

theoretical and technical foundation of lightweight design in the manufacturing process.

Meanwhile, due to the application of universal finite element analysis software in clothing design,

many clothing problems have been solved. Horiba et al. [3] combined finite element analysis

software with other clothing problem solving tools to effectively estimate clothing problems,

including shear stress, heat conduction, pressure, temperature, etc. By combining different solving

tools and software, the optimal method can be found to estimate clothing problems, and the

impact of clothing structure and material characteristics on clothing performance can be better

understood. Hu [4] proposed a design framework for a component-based intelligent clothing

modeling CAD system. By dividing clothing into different components, each component becomes a

relatively independent design unit. Huang [5] applied the psychological recognition system to

clothing design and made algorithm improvements. By developing a clothing recognition structure

system with different psychological electrical signals, the design and construction of intelligent

clothing were carried out. Through the design research in this article, the test evaluation results

were used to experience the effectiveness of clothing design. Lee et al. [6] established and

optimized the manufacturing process of intelligent sports underwear using various automated

machines. Li et al. [7] conducted clothing design development on the Internet and artificial

intelligence. The study explores some challenges in design, raw materials, and supply chain

management from the perspective of the clothing industry chain. Through personalized design of

medical and clothing, it has analyzed the hot research areas of intelligent clothing. Lashin et al. [8]

designed a fuzzy design system for artificial intelligence. The system sets fuzzy logic for different

Computer-Aided Design & Applications, 20(S13), 2023, 111-123

113

optimized stores by controlling influencing factors. Design through the color, lighting, and logo of

the product, aiming to achieve more perfect customer optimization and logical setting. Linet et al.

[9] conducted an open online graphic training access software, which showed that the basic

development mode of CAD open software requires playback measurement tools and video editing.

This study can help teachers who face challenges in accessing educational software for specific

disciplines, guiding them on how to use affordable alternative software. Produce teaching materials

such as screen projection for teaching CAD concepts, such as Coreldraw. The production of

patterns is the beginning of the clothing design cycle. Pattern making is a mature technology that

requires technical ability, flexibility in design interpretation, and a realistic understanding of

clothing structure. Moniruzzaman and Oishe [10] designed and developed a technique called

planar pattern design to construct various types of patterns. Ninga [11] elaborated on the current

application status and the sorting in teacher work tasks, using typical clothing product types as

carriers and enterprise actual tasks as carriers, using classroom teaching methods, emphasizing

the consistency between students' learning and actual work. Starting from students' practical

teaching. Panneerselvam and Prakash [12] perfectly edit the pattern outline according to the

required shape in the graphic design of jacquard fabrics. The weaving markers that need to be

applied should control long floats and avoid using them in places where there are no longer floats

to maintain the perfect editing contour. In manual graphic design, the designer decides on the

type of woven marker used to control the float and applies it to the selected part. Sayem [13]

used computer programs to calculate the similarity between actual pants and virtual clothing. This

can be achieved by using computer vision techniques such as calculating similarity matrices or

Euclidean distances. Virtual 3D CLO (Computer Integrated Geometry Processing) programs can be

used to compare the similarities between actual pants and virtual pants. Souza [14] aims to

explore the application of intelligent technology in the clothing and accessories industry for people

with disabilities. It describes the intelligent technologies in ves tiles devices aimed at helping

patients recover through a systematic review of the database. Won and Lee [15] imported image

data of actual and virtual pants into a computer. And use appropriate software to convert them

into 3D geometric shapes. This can be achieved by using techniques such as triangular meshes or

surface meshes. Xin et al. [16] conducted an analysis of the application of nanotechnology

materials in textile and clothing. By further analyzing the market and consumption, safe and

intelligent design of nanomaterials can be carried out. Its research aims to apply safety

intelligence under nano clothing. Promote the innovative development of textile and clothing,

further adapt to the new needs of the market and consumers, and play a positive role.

3 RESEARCH METHOD

3.1 Pattern Matching Algorithm for Clothing CAD Technology

Due to the rise of computer technology, a large number of ethnic clothing patterns have emerged

on the internet, which are diverse and often unable to be classified. Using web crawler and other

technologies can easily search and download images in the network, that is, the problem of

obtaining national costume patterns in the network is easy to solve. How to extract patterns with

unique meanings is particularly important. Especially after segmenting ethnic patterns, there will

be many repetitive and flawed patterns. Extracting feature patterns using matching algorithms is

an effective method.

Extracting pattern texture features for matching, as this feature has the characteristics of low

dimensionality, good matching effect, and strong stability. The method of structure is that the

pattern is composed of texture elements with a "repetitive" spatial organization structure and

arrangement rules. Representative methods include syntactic texture description and digital

morphology. SIFT (Scale Invariant Feature Transform) is a pattern feature descriptor. This

descriptor describes the local features of the pattern, has scale invariance, and can detect key

points of multi-scale patterns, playing an important role in pattern matching.

Computer-Aided Design & Applications, 20(S13), 2023, 111-123

114

Gaussian blur, also known as Gaussian smoothing, is a pattern filter. It uses Gaussian function

to calculate the blur template. The two-dimensional scale space of a pattern is defined as follows:

( ) ( ) ( )

yxIyxGyxL ,,,,, =



(1)

( )

( ) ( )

( )

2

22

2

2/2/

2

1

,,







nymx

eyxG

−+−

=

(2)

( )



,, yxL

represents the scale space of the pattern,

( )



,, yxG

is a Gaussian function with

variable scale, "



" is a convolution operator, and

( )

yx,

is the size of spatial coordinates that

determines the smoothness of the pattern. The larger



is, the smoother the pattern is, and the

smaller



is, the clearer the pattern is. The pyramid model is shown in Figure 1 below:

Figure 1: Gaussian pyramid model.

Because the second derivative Laplacian is very sensitive to image noise, it is usually not used

directly. Instead, the influence of the second derivative Laplacian on the increase of image noise,

such as Gaussian Laplacian, is alleviated by introducing Gaussian functions. The definition of

Laplacian operator is shown in Formula (3):

( )

( ) ( )

2

,,

,

y

yxf

x

yxf



+



=

(3)

Gauss Laplacian operator is defined as shown in formula (4):

( )

( ) ( )

2

22

2

4

222

2

2,,

,



yx

e

yx

y

xxG

x

xxG

yxG

+













−+

=



+



=

(4)

The value



is the standard deviation of Gaussian function

( )

yxG ,

.

( )

CZZF +=

2

(5)

CZ,

is plural here. Therefore, in the actual calculation and programming, the corresponding

real part and imaginary part are used to replace

CZ,

here.

Computer-Aided Design & Applications, 20(S13), 2023, 111-123

115

At present, the clothing CAD system includes the following content: (1) Style design system:

the application of clothing shape design and color. It mainly consists of functions such as line

drawing, color pattern filling, effect polishing, and effect image printing and output. (2) The paper

pattern structure design system refers to the design of clothing plane structure. The main process

involves selecting design methods, determining specifications and standards, analyzing and

calculating data, analyzing and determining structural elements, and designing and drawing paper

patterns. (3) Template scaling system: also known as grading, push board, etc. The main process

includes basic paper sample input, design of grading rules, and drawing of grading quantity input

paper sample scaling diagram. (4) Layout diagram design system: connected to the automatic

cutting bed system, providing an important basis for the subsequent process of cutting. The main

process includes design of bed separation scheme, preprocessing of discharge data, selection of

discharge scheme, and drawing of discharge diagram. Due to the limited number of texture

features and fixed angle rotation transformations in clothing patterns, it is difficult to match the

patterns. Therefore, this article proposes a further improvement plan to extract the main direction

vector of the pattern shape context descriptor, and add rotation invariance to the algorithm

through rotation matching of the direction vector to adapt to the matching application of ethnic

patterns.

The required rotation angle of the pattern is determined by matching the pattern direction

vector, and the matching cost is calculated after the pattern is rotated. Then, the matching results

of the improved algorithm on the national ornamentation pattern database and the traditional

shape database are shown. The matching process is shown in Figure 2.

Figure 2: Matching method flow.

Computer-Aided Design & Applications, 20(S13), 2023, 111-123

116

The unstable value of the extremum can be defined as: let the region

i

C

be an extremum region,

and

( ) ( )

+=

+ iiii

CCCCS ,,,

1



is the branch node of the root node

i

C

. The unstable value of

i

C

is defined as:

( )

i

ii

i

C

CC

Cv

−

=

+

(5)

Where

i

C

represents the number of pixels in

i

C

. If the unstable value of the extreme value area

i

C

is smaller, that is, more stable, than that of its parent node

1−i

C

and child node

1+i

C

, this

extreme value area

i

C

is called the maximum stable extreme value area.

In this paper, we choose the combination of energy, contrast and homogeneity as the rule.

Energy, measured by ASM (AngularCondMoment), uses

( )

jiC ,

itself as its own weight.

( )



= =

==

0 0

2

,,

i j

jiCASMASMEnergy

(6)

The weights of contrast and homogeneity features are affected by the distance between pixels,

such as formula (7).

( ) ( )

jiCjiContrast

i j

,

0 0

2



= =

−=

(7)

Among them, the more uniform the distribution of

( )

jiC ,

, the smaller the energy characteristic,

and when all terms are equal, the characteristic value is the smallest.

The value in the initial direction vector is not 0 or 1, but the number of feature points in each

interval. If all feature points are set to

120=P

, then the extracted direction vector is

( )

0,130,4,11O

, and

O

is shifted to get

( )

1,0,0,0

'

O

. It is necessary to calculate the Euclidean

distance between

O

and

21

,QQ

to judge the most appropriate direction vector. As shown in the

following formula (8).

( )



−=

2

oOd

(8)

3.2 DL-Based Pattern Generation

The 3D clothing CAD system is used as a display and pattern design tool for clothing 3D. It mainly

has the following functions: linking fabric models with objective test data, providing realistic fabric

drape models. 3D to ZD flat unfolding algorithm and providing automatic plate making.

Automatically push gears using a scaled manikin. Dressing technology combines traditional design

patterns into clothing, which is then observed on a 3D mannequin. Utilizing personalized 3D

human body models for customized clothing design, namely MTM technology for clothing. Figure 3

shows a typical CNN (Convolutional Neural Network) structure.

In that convolution lay, each feature graph has a convolution kernel with the same size, and

each feature graph in the convolution layer is convolve on the feature graph input in the previous

layer by different convolution kernels, then the correspond elements are multiplied and added,

then an offset is added, and finally, the activation function is used to convert it into nonlinear

output.

Computer-Aided Design & Applications, 20(S13), 2023, 111-123

117

Figure 3: The basic structure of CNN.

CNN's output layer needs to classify images, so it needs classifiers. Common classifiers include

Softmax and Sigmoid. The max function is calculated as formula (9).

( )



=

k

T

i

T

i

x

ip

1

exp



(9)

In the original CAD image description generative model based on adaptive attention mechanism,

we improved the language decoder because of the simple structure of the language decoder. Our

new model uses a two-layer LSTM (short - and long-term memory) network in the decoding part.

Perform the required mapping through sufficient depth and nonlinear transformation, instead of

placing a simple single hidden layer multilayer perceptron on the top of the decoder in the original

language decoding model. The network structure of the model proposed in this article is shown in

Figure 4.

Figure 4: The network structure of the model.

The encoder is implemented by Transformer structure. Given a set of image features

I

extracted

from the input image, the permutation invariant coding of

X

can be obtained by

attentionself −

used in Transformer:

( ) ( )

V

d

QK

softVKQattentionIattentionself

T













==− max,,

(10)

Q

is the query vector,

VK,

is a set of corresponding data pairs;

d

is the scaling factor.

Computer-Aided Design & Applications, 20(S13), 2023, 111-123

118

There is no need to establish the graph structure relationship of words for iterative operation.

Therefore, the generation description text model of this paper adopts TF-IDF to extract keywords,

and the calculation formula is:

( )

title

YTF-IDFtopic =

(11)

The two layers of LSTMs in the decoder are standard LSTM units, and the general formula for

calculating the hidden layer state of lstm units is as follows:

( )

1

,

−

=

ttt

hxLSTMh

(12)

Where

t

x

represents the input of the LSTM unit at the current time

t

, and

1−t

h

represents the

hidden layer state of the LSTM unit at the previous time.

4 ANALYSIS AND DISCUSSION OF RESULTS

According to the requirements of virtual clothing design, clothing designers can use the 3D V

Resign design system to directly translate or rotate along the y-axis or z-axis using the system's

internal dialog box. Simultaneously utilizing the Opengl graphics library developed by SG I

company, utilizing its graphics functionality and cross platform capabilities. By combining computer

vision methods, read the three-dimensional lattice data of the human body from several existing

images. Subsequently, the 3D human body model surface fitting function is utilized. By combining

digital cameras, rotating platforms, and the rational application of dense lighting, a complete 3D

human clothing design software model is constructed. The experimental environment of this paper

is Intel(R)Core(TM)i5 CPU with 2GB memory. MATLAB is used to realize this algorithm. The images

of ethnic patterns of six ethnic minorities were collected, and a data set of ethnic images with

1000 images was formed through data expansion, of which 800 were used as training sets and

200 were used as testing sets.

In order to recognize the characters in the scene, this model is trained for each type of

characters on the picture to detect the potential character regions and select the region with the

highest detection score as the character recognition result. The results are shown in Table 1 and

Figure 5, and compared with LSTM method.

Category

LSTM

our

Clear

75.432

87.212

Complex background

83.251

83.549

Multicolor character

76.127

88.03

Uneven illumination

76.271

84.905

Low contrast

79.067

84.5

Blurred

78.232

86.131

Severe deformation

76.975

83.306

Table 1: Comparison of recognition rate.

The average recognition rate of this model on data sets is 85.3761%, that of LSTM is 77.9079%,

and that of various characters is increased by 3.06%-10.18% respectively. Moreover, the influence

of character classification on the recognition rate of this model is smaller than that of LSTM,

especially for pictures with uneven illumination, which proves that the robustness of this model is

stronger than that of LSTM. In order to better evaluate the recognition performance of this model,

we evaluate the performance of the algorithm on data sets. The experimental results are shown in

Figure 6.

Computer-Aided Design & Applications, 20(S13), 2023, 111-123

119

Figure 5: Comparative statistical chart of recognition rate.

Figure 6: Character recognition rate of several days in different candidate areas.

The results show that the average recognition rate of this model on the data set increases by

3.15%, and with the increase of candidate character regions, this model can achieve a higher

recognition rate than LSTM. Aiming at the dimensionality reduction of high-dimensional sparse

coding descriptors, effective information is saved, and the recognition accuracy is improved.

Compared with gradient features, sparse coding has stronger expressiveness to characters.

Time overhead is an angle to measure the algorithm, and the ideal algorithm is to reduce time

overhead while pursuing high accuracy. For real-time recognition algorithms, time overhead is a

very important aspect that needs to be optimized. Table 2 and Figure 7 show the time cost

comparison of different algorithms.

Image

AlexNet

VGGNet

ResNet

LSTM

Our

Computer-Aided Design & Applications, 20(S13), 2023, 111-123

120

sample

1

2.629

2.61

5.188

4.495

0.864

2

1.485

3.492

3.717

4.98

0.975

3

3.384

2.83

2.71

3.506

0.612

4

3.748

3.269

3.776

4.031

1.292

5

1.863

3.813

3.329

2.788

1.098

6

2.543

2.282

3.286

2.79

0.665

7

1.812

3.673

4.995

4.473

0.884

8

1.686

2.921

4.563

3.569

1.397

9

2.728

2.361

5.013

2.597

0.76

10

2.2

2.481

4.336

4.202

1.251

11

3.552

4.123

3.627

3.623

1.077

12

2.948

3.941

4.986

4.267

0.89

13

3.49

2.284

4.416

5.104

0.666

14

3.025

3.665

3.755

4.599

0.484

15

1.734

2.435

2.686

3.894

0.915

Table 2: Time cost comparison (seconds).

Figure 7: Time expenditure trend chart.

It can be seen that the time cost of this algorithm is the lowest, and the time cost of ResNet is the

highest, which is about three times that of the former. Net ranked second, and VGGNet ranked

third. Time overhead is directly related to the length of feature vector, and the longer the length,

the greater the time overhead. In this paper, after extracting features and effectively fusing them,

the algorithm adopts dimension reduction technology, which greatly reduces the time cost.

From the above matching results, it can be seen that the improved algorithm in this paper can

basically find out the matching pattern with four angular rotations similar to the original pattern.

The average accuracy rate of the whole national pattern search is shown in Figure 8 and Figure 9

below.

Computer-Aided Design & Applications, 20(S13), 2023, 111-123

121

Figure 8: The algorithm in this paper matches the traditional dress primitives.

Figure 9: The traditional method matches the traditional dress primitives.

In this paper, the algorithm greatly improves the accuracy of pattern matching with four angular

rotation changes, and the precision rate is increased by 56.127%. Good results have been

achieved in the matching of national ornamentation patterns, which can match similar patterns

that can't be found in traditional shape context methods due to rotation transformation.

From the highest precision of single pattern, the highest precision of the traditional algorithm is

only 70.066%, while for some ethnic patterns, the highest precision of this improved algorithm can

reach 100%.

We replace the image encoder in the Adaptive Attention model with the ResNet-101 network,

and train it again. During the training, only the parameters of the sentence decoding model are

trained, and the CNN part is not fine-tuned, so as to achieve fair comparison. Table 3 shows the

Computer-Aided Design & Applications, 20(S13), 2023, 111-123

122

comparative experimental results of Adaptive Attention model and our new model on ethnic image

data sets.

Evaluation criteria

Adaptive Attention

our

BLEU

0.842

0.831

METEOR

0.9

0.95

ROUGE

0.975

1.483

CIDEr

1.413

1.65

Table 3: Comparative experimental results.

It can be seen that the performance of our improved model is better than that of the original

Adaptive Attention model, which verifies the effectiveness of our double-layer LSTM language

decoding model. The language decoding model based on double-layer LSTM can perform more

nonlinear transformations on image information and language information, thus improving the

semantic expression ability of the generated description sentences.

5 CONCLUSION

By utilizing 3D human body measurement technology and virtual reality technology, a fully sized

and realistic human body model and clothing mannequin can be established in a computer.

Designers can rotate the model from various angles and design the direction of light around the

stage, creating an immersive feeling. This article proposes an application of DL intelligent vision

computer-aided technology in clothing culture. Through network transmission, designers can use

virtual reality technology to build 3D models for customers according to the size of their

corresponding parts input by customers. And implement personalized design with human absence

on its model. After virtual sewing with a computer and communicating with customers about the

fitting effect through the internet, the entire design of the work is ultimately completed. In this

article, the algorithm greatly improves the accuracy of pattern matching with changes in corner

rotation, with an accuracy increase of 56.127%. Good results have been achieved in the matching

of ethnic decorative patterns. From the perspective of the highest accuracy of a single pattern, the

traditional algorithm has a maximum accuracy of only 70.066%, while for some ethnic patterns,

the improved algorithm can achieve a maximum accuracy of 100%.

6 ACKNOWLEDGEMENTS

This work was supported by General project of art science of National Social Science Foundation, A

study on the pattern drawing and construction thought of “Different people wear different

uniforms” of men’s clothing in the Northern Song Dynasty (No.: 18BG112).

Xianyu Wang, https://orcid.org/0000-0003-2534-3205

Xiaoguang Sun, https://orcid.org/0009-0005-2918-7184

REFERENCES

[1] Abdurahimovna, U.-F.: Achieving Educational Effectiveness Using ICT Tools in the

Development of Competence in the Design and Modeling of Clothing in the Organization of

Specialized Disciplines, Journal of Asian Multicultural Research for Educational Study, 2(1),

2020, 001-004. https://doi.org/10.47616/jamres.v2i1.100

[2] Du, H.; Jiang, Q.; Xiong, W.: Computer-aided optimal design for flexible cable in aerospace

products based on dynamic analogy modeling, Scientific Reports, 12(1), 2022, 5833.

https://doi.org/10.1038/s41598-022-09880-9

Computer-Aided Design & Applications, 20(S13), 2023, 111-123

123

[3] Horiba, Y.; Amano, T.; Inui, S.; Yamada, T.: Proposal of method for estimating clothing

pressure of tight-fitting garment made from highly elastic materials: hybrid method using

apparel CAD and finite element analysis software, Journal of Fiber Science and Technology,

77(2), 2021, 76-87. https://doi.org/10.2115/fiberst.2021-0006

[4] Hu, L.: Design and implementation of a component-based intelligent clothing style CAD

system, Computer-Aided Design and Applications, 18(S1), 2020, 22-32.

https://doi.org/10.14733/cadaps.2021.S1.22-32

[5] Huang, M.: Application of Behavioral Psychology in Clothing Design from The Perspective of

Big Data, Applied Artificial Intelligence, 37(1), 2023, 2194118.

https://doi.org/10.1080/08839514.2023.2194118

[6] Lee, S.; Rho, S.-H.; Lee, S.; Lee, J.; Lee, S.-W.; Lim, D.; Jeong, W.: Implementation of an

automated manufacturing process for smart clothing: The case study of a smart sports bra,

Processes, 9(2), 2021, 289. https://doi.org/10.3390/pr9020289

[7] Li, Q.; Xue, Z.; Wu, Y.; Zeng, X.: The status quo and prospect of sustainable development of

smart clothing, Sustainability, 14(2), 2022, 990. https://doi.org/10.3390/su14020990

[8] Lashin, M.-M.; Khan, M.-I.; Khedher, N.-B.; Eldin, S.-M.: Optimization of Display Window

Design for Females’ Clothes for Fashion Stores through Artificial Intelligence and Fuzzy

System, Applied Sciences, 12(22), 2022, 11594. https://doi.org/10.3390/app122211594

[9] Linet, M.; Chipo, C.; Felisia, C.: Online Instructional Material for Computer Aided Garment

Pattern Making Training in Colleges: A Case Study of Zimbabwe, International Journal of

Costume and Fashion, 21(1), 2021, 54-66. https://doi.org/10.7233/ijcf.2021.21.1.054

[10] Moniruzzaman, R.-A.-A.; Oishe, S.-H.: An approach to design solutions for garments using a

CAD system, J Textile Eng Fashion Technol, 8(4), 2022, 145-148.

https://doi.org/10.15406/jteft.2022.08.00313

[11] Ninga, Z.: Research on the application of computer aided Design in clothing Design teaching

in higher vocational colleges, Turkish Journal of Computer and Mathematics Education

(TURCOMAT), 12(3), 2021, 4817-4821. https://doi.org/10.17762/turcomat.v12i3.1985

[12] Panneerselvam, R.-G.; Prakash, C.: Study on the float-control weaves application algorithms

of computer-aided jacquard graph designing for different figured fabrics, The Journal of The

Textile Institute, 114(1), 2023, 88-100. https://doi.org/10.1080/00405000.2021.2023956

[13] Sayem, A.-S.-M.: Digital fashion innovations for the real world and metaverse, International

Journal of Fashion Design, Technology and Education, 15(2), 2022, 139-141.

https://doi.org/10.1080/17543266.2021.1938701

[14] Souza, G.-S.: Overview of intelligent clothing and accessories technology system for the

disabled, Wearable Technology, 2(1), 2021, 51-59. http://dx.doi.org/10.54517/wt.v2i1.1673

[15] Won, Y.; Lee, J.-R.: A Study on the Comparison of Fit Similarity Between the Actual and

Virtual Clothing According to the Pants Silhouette, Fashion & Textile Research Journal, 23(6),

2021, 826-835. https://doi.org/10.5805/SFTI.2021.23.6.826

[16] Xin, Y.; Zhang, D.; Qiu, G.: Application of nanomaterials in safety intelligent clothing design,

Integrated Ferroelectrics, 216(1), 2021, 262-275.

https://doi.org/10.1080/10584587.2021.1911293