Artificial Intelligence on Fashion and Textiles

Advances in Intelligent Systems and Computing 849

Wai Keung Wong Editor

Artificial

Intelligence on

Fashion and Textiles

Proceedings of the Artificial

Intelligence on Fashion and Textiles

(AIFT) Conference 2018, Hong Kong,

July 3–6, 2018

Advances in Intelligent Systems and Computing

Volume 849

Series editor

Janusz Kacprzyk, Polish Academy of Sciences, Warsaw, Poland

e-mail: [email protected]

The series “Advances in Intelligent Systems and Computing” contains publications on theory,

applications, and design methods of Intelligent Systems and Intelligent Computing. Virtually all

disciplines such as engineering, natural sciences, computer and information science, ICT, economics,

business, e-commerce, environment, healthcare, life science are covered. The list of topics spans all the

areas of modern intelligent systems and computing such as: computational intelligence, soft computing

including neural networks, fuzzy systems, evolutionary computing and the fusion of these paradigms,

social intelligence, ambient intelligence, computational neuroscience, artificial life, virtual worlds and

society, cognitive science and systems, Perception and Vision, DNA and immune based systems,

self-organizing and adaptive systems, e-Learning and teaching, human-centered and human-centric

computing, recommender systems, intelligent control, robotics and mechatronics including

human-machine teaming, knowledge-based paradigms, learning paradigms, machine ethics, intelligent

data analysis, knowledge management, intelligent agents, intelligent decision making and support,

intelligent network security, trust management, interactive entertainment,Web intelligence and multimedia.

The publications within “Advances in Intelligent Systems and Computing” are primarily proceedings

of important conferences, symposia and congresses. They cover significant recent developments in the

field, both of a foundational and applicable character. An important characteristic feature of the series is

the short publication time and world-wide distribution. This permits a rapid and broad dissemination of

research results.

Advisory Board

Chairman

Nikhil R. Pal, Indian Statistical Institute, Kolkata, India

e-mail: [email protected]

Members

Rafael Bello Perez, Universidad Central “Marta Abreu” de Las Villas, Santa Clara, Cuba

e-mail: [email protected]

Emilio S. Corchado, University of Salamanca, Salamanca, Spain

e-mail: [email protected]

Hani Hagras, University of Essex, Colchester, UK

e-mail: [email protected]

László T. Kóczy, Széchenyi István University, Győr, Hungary

e-mail: [email protected]

Vladik Kreinovich, University of Texas at El Paso, El Paso, USA

e-mail: [email protected]

Chin-Teng Lin, National Chiao Tung University, Hsinchu, Taiwan

e-mail: [email protected]

Jie Lu, University of Technology, Sydney, Australia

e-mail: [email protected]

Patricia Melin, Tijuana Institute of Technology, Tijuana, Mexico

e-mail: [email protected]

Nadia Nedjah, State University of Rio de Janeiro, Rio de Janeiro, Brazil

e-mail: [email protected]

Ngoc Thanh Nguyen, Wroclaw University of Technology, Wroclaw, Poland

e-mail: [email protected]

Jun Wang, The Chinese University of Hong Kong, Shatin, Hong Kong

e-mail: [email protected]

More information about this series at http://www.springer.com/series/11156

Wai Keung Wong

Editor

Artificial Intelligence

on Fashion and Textiles

Proceedings of the Artificial Intelligence

on Fashion and Textiles (AIFT) Conference

2018, Hong Kong, July 3–6, 2018

123

Editor

Wai Keung Wong

Institute of Textiles and Clothing

The Hong Kong Polytechnic University

Hunghom, Hong Kong

ISSN 2194-5357 ISSN 2194-5365 (electronic)

Advances in Intelligent Systems and Computing

ISBN 978-3-319-99694-3 ISBN 978-3-319-99695-0 (eBook)

https://doi.org/10.1007/978-3-319-99695-0

Library of Congress Control Number: 2018952621

© Springer Nature Switzerland AG 2019

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part

of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,

recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission

or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar

methodology now known or hereafter developed.

The use of general descriptive names, registered names, trademarks, service marks, etc. in this

publication does not imply, even in the absence of a specific statement, that such names are exempt from

the relevant protective laws and regulations and therefore free for general use.

The publisher, the authors and the editors are safe to assume that the advice and information in this

book are believed to be true and accurate at the date of publication. Neither the publisher nor the

authors or the editors give a warranty, express or implied, with respect to the material contained herein or

for any errors or omissions that may have been made. The publisher remains neutral with regard to

jurisdictional claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Switzerland AG

The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Contents

A Clothing Recommendation System Based on Expert Knowledge .... 1

Tao Yang, Jiao Feng, Jie Chen, Chunyan Dong, Youqun Shi and Ran Tao

Coordinated Optimization of Production and Delivery Operations in

Apparel Supply Chains Using a Hybrid Intelligent Algorithm ........ 9

Zhaoxia Guo, Jingjie Chen, Guangxin Ou and Haitao Liu

Intelligent Cashmere/Wool Classification with Convolutional Neural

Network ................................................. 17

Fei Wang, Xiangyu Jin and Wei Luo

Yarn Quality Prediction for Spinning Production Using the Improved

Apriori Algorithms......................................... 27

Xianhui Zeng and Pengcheng Xing

Everybody Immersive Fashion_Human–Computer

Interaction in VR .......................................... 37

Kate Kennedy

Fabric Defect Detection Based on Faster RCNN .................. 45

Bing Wei, Kuangrong Hao, Xue-song Tang and Lihong Ren

Woven Light: An Investigation of Woven Photonic Textiles ......... 53

Lan Ge, Jeanne Tan, Richard Sorger and Ziqian Bai

A Piezoelectric Energy Harvester for Wearable Applications......... 61

Wenying Cao, Weidong Yu and Wei Huang

Surrogate-Based Modeling and Optimization of the Bleach Washing

for Denim Fabrics ......................................... 69

Wenbo Ke, Jie Xu, Ming Yang and Changhai Yi

Costume Expert Recommendation System Based on Physical

Features ................................................. 77

Aihua Dong, Qin Li, Qingqing Mao and Yuxuan Tang

v

Cognitive Characteristics Based Autonomous Development of

Clothing Style ............................................. 87

Jiyun Li and Xiaodong Zhong

Fabric Identification Using Convolutional Neural Network .......... 93

Xin Wang, Ge Wu and Yueqi Zhong

Discrete Hashing Based Supervised Matrix Factorization

for Cross-Modal Retrieval ................................... 101

Baodong Tang, Xiaozhao Fang, Shaohua Teng, Wei Zhang

and Peipei Kang

Sparse Discriminant Principle Component Analysis ................ 111

Zhihui Lai, Mangqi Chen, Dongmei Mo, Xingxing Zou and Heng Kong

New Product Design with Popular Fashion Style Discovery Using

Machine Learning ......................................... 121

Jiating Zhu, Yu Yang, Jiannong Cao and Esther Chak Fung Mei

Challenges in Knitted E-textiles ............................... 129

Amy Chen, Jeanne Tan, Xiaoming Tao, Philip Henry and Ziqian Bai

The CF+TF-IDF TV-Program Recommendation .................. 137

Li Yan, Cui Jinrong, Xin Liu, Yu JiaHao and He Mingkai

Minimize the Cost Function in Multiple Objective Optimization by

Using NSGA-II ............................................ 145

Hayder H. Safi, Tareq Abed Mohammed and Zena Fawzi Al-Qubbanchi

Two-Layer Mixture Network Ensemble for Apparel Attributes

Classification ............................................. 153

Tianqi Han, Zhihui Fu and Hongyu Li

3D Digital Modeling and Design of Custom-Fit Functional

Compression Garment ...................................... 161

Rong Liu and Bo Xu

Fine-Grained Apparel Image Recognition Based on Deep Learning ... 171

Jia He, Xi Jia, Junli Li, Shiqi Yu and Linlin Shen

Learning a Discriminative Projection and Representation for Image

Classification ............................................. 179

Zuofeng Zhong, Jiajun Wen, Can Gao and Jie Zhou

Fashion Outfit Style Retrieval Based on Hashing Method ........... 187

Yujuan Ding and Wai Keung Wong

Supervised Locality Preserving Hashing......................... 197

Xiao Zhou, Zhihui Lai and Yudong Chen

vi Contents

Co-designing Interactive Textile for Multisensory Environments ...... 205

H. Y. Kim, J. Tan and A. Toomey

Parametric Stitching: Co-designing with Machines ................ 213

Jenny Underwood

Live: Scape BLOOM: Connecting Smart Fashion

to the IoT Ecology ......................................... 221

Caroline McMillan

Traps in Multisource Heterogeneous Big Data Processing ........... 229

Yan Liu

Convolutional Neural Networks for Finance Image Classification ..... 237

Xingjie Zhu, Yan Liu, Xingwang Liu and Chi Li

Rough Possibilistic Clustering for Fabric Image Segmentation ....... 247

Jie Zhou, Can Gao and Jia Yin

Fashion Meets AI Technology ................................ 255

Xingxing Zou, Wai Keung Wong and Dongmei Mo

Fashion Style Recognition with Graph-Based Deep Convolutional

Neural Networks .......................................... 269

Cheng Zhang, Xiaodong Yue, Wei Liu and Can Gao

Fabric Defect Detection with Cartoon–Texture Decomposition ....... 277

Ying Lv, Xiaodong Yue, Qiang Chen and Meiqian Wang

Fabric Texture Removal with Deep Convolutional Neural Networks ... 285

Li Hou, Xiaodong Yue, Xiao Xiao and Wei Xu

Optimal Gabor Filtering for the Inspection of Striped Fabric ........ 291

Le Tong, Xiaoping Zhou, Jiajun Wen and Can Gao

Robust Feature Extraction for Material Image Retrieval in Fashion

Accessory Management ..................................... 299

Yuyang Meng, Dongmei Mo, Xiaotang Guo, Yan Cui, Jiajun Wen

and Wai Keung Wong

Woven Fabric Defect Detection Based on Convolutional Neural

Network for Binary Classification ............................. 307

Can Gao, Jie Zhou, Wai Keung Wong and Tianyu Gao

Complex Textile Products and Reducing Consumer Waste .......... 315

Colin Gale

A Fast Parallel and Multi-population Framework with

Single-Objective Guide for Many-Objective Optimization ........... 321

Haitao Liu, Weiwei Le and Zhaoxia Guo

Contents vii

Multiple Criteria Group Decision-Making Based on Hesitant Fuzzy

Linguistic Consensus Model for Fashion Sales Forecasting .......... 329

Ming Tang and Huchang Liao

Probabilistic Linguistic Linear Least Absolute Regression

for Fashion Trend Forecasting ................................ 337

Lisheng Jiang, Huchang Liao and Zhi Li

Author Index................................................ 347

viii Contents

A Clothing Recommendation System

Based on Expert Knowledge

Tao Yang, Jiao Feng, Jie Chen, Chunyan Dong, Youqun Shi and Ran Tao

Abstract Through summarizing expert experience and knowledge of clothing, the

clothing recommendation system is developed based on a kind of clothing recom￾mendation method. According to color matching rules, this method has refined the

six factors that affect the customer’s choice of clothing, establish the clothing knowl￾edge base and clarify the recommendation rules. Considering the characteristics of

the customers and the selection criteria, this system can make personalized clothing

recommendation scheme for customers and ensure the rationality of the recommen￾dation results.

Keywords Expert system · Cloth recommendation · Knowledge base

1 Introduction

With the increase of individualized wearable consciousness, clothing is not only the

basic living demand, but also the important carrier to enhance self-taste and image.

T. Yang (B) · J. Feng · J. Chen · C. Dong · Y. Shi · R. Tao

Donghua University, Shanghai 200051, China

e-mail: [email protected]

J. Feng

e-mail: [email protected]

J. Chen

e-mail: [email protected]

C. Dong

e-mail: [email protected]

Y. Shi

e-mail: [email protected]

R. Tao

e-mail: [email protected]

© Springer Nature Switzerland AG 2019

W. K. Wong (ed.), Artificial Intelligence on Fashion and Textiles,

Advances in Intelligent Systems and Computing 849,

https://doi.org/10.1007/978-3-319-99695-0_1

1

2 T. Yang et al.

In the daily shopping, the various clothes on the e-commerce platform often make

customers regret consumption [1].

Based on the above background, a personalized clothing recommendation system

based on expert system has been designed and developed in this paper, which aims

to guide consumers to choose the suitable clothing on the perspective of professional

match. The main work is the application of clothing recommendation and match

experience provided by experts in personalized cloth recommendation. We have

refined six factors affecting customer dress, transformed the thinking mode of experts

into the electronic knowledge base and recommendation process that computer can

handle.

2 Clothing Recommendation Knowledge Base

The knowledge base in the clothing recommendation system mainly refers to the set

of rules used by the system runtime, including the data information corresponding

to the rules and the storage mode that rules can be transformed and processed by

computers after summarizing experts’ experience [2].

2.1 Customer Characteristics and Clothing Elements

Recommendation based on expert rules uses experts’ knowledge which can map

customers’ needs to product features and take customers’ attributes as the main

consideration [3]. The recommendation system depends on two parts of customers

and clothing, as shown in Fig. 1. Integrating with experts’ years of experience to

identify customer characteristics and clothing elements is the basis of knowledge

base establishment and clothing recommendation realization.

Customer Characteristics Extraction. When experts design customer’s image,

the first consideration is the customer’s skin color which influences the customer’s

suitable color range. The second point is to consider the body type. The correct use

of clothing version can make up body defects. In order to clarify the customer’s

preferences, the style factor is considered. Each style has a corresponding theme

color. Finally, the recommended results are given based on customer demand for

clothing categories and specific colors [4].

The customers’ information we collect are divided into two parts: objective factors

and subjective factors, as shown in Fig. 1. Objective factors refer to the basic attributes

of customers, including age, height, skin color, and measurements of chest, waist,

and hips. Height and weight measurements are used to determine the customer’s

body type. Subjective elements are updated as the preferences of customer change,

including style, preference color and category. Based on the above-mentioned rules,

customer factors can be the following four options:

A Clothing Recommendation System Based on Expert Knowledge 3

Fig. 1 Customer characteristics and clothing elements

• Age range. In addition to a small number of basic styles of clothing, clothing has

its age range which is suitable for customers, the age of the customer determines

the choice of clothing A4.

• Version range. Clothing version can help customers highlight the advantages of

figure while concealing defects, so obtaining customer size data means that system

can choose the appropriate range of clothing version A2.

• Color range. In customer characteristics, the three factors including skin color,

style, and preference color are related to color. Skin color influences the color range

of the customer. Style determines the theme color. Preference color defines color

selection range. The combination of the three factors can generate the customer

clothing’s best color range A3.

• Category range. Customer’s demand for a particular category is the most intuitive

clothing conditions which can obtain the category range A1.

Clothing Elements Extraction. Clothing description is inseparable from the cloth￾ing name, fabric, composition, brand and other information, shown in Fig. 1, we

consider this information as the basic elements. When the system recommends the

clothing, the basic information is far from enough. In order to be consistent with

customer factors, we join the four extended elements which contain category, age

range, the main color, and clothing version. Expanded elements need to meet the

system-specific classification and data requirements [5]. As an example, the primary

classification of clothing is the cloth of upper body. The second classification for

this primary classification is T-shirt, shirt, sweater and so on. Main color refers to

the largest proportion of fabric color. The clothing version refers to the outer con￾tour of the clothing. When matching the clothing for the customer, the customer’s

range of choice for clothing and clothing elements are be consistent one-to-one [6].

The clothing’s category elements A1 correspond to the customer’s category range.

The clothing’s suitable age range elements A4 corresponds to the customer’s age

4 T. Yang et al.

range A4. The clothing’s main color elements A3 corresponds to the customer’s

color range A3. The clothing’s version elements A2 corresponds to the customer’s

version range A2.

2.2 Knowledge Base

Clothing knowledge base consists of three parts, including element library, color

library, and recommendation rules algorithm, as shown in Fig. 2. The element library

is divided into customer elements and clothing elements which store the classifica￾tion of each factor and the corresponding color range. Color library includes the

correspondence of Pantone and RGB, color similarity calculation method, and basic

color expansion. The rule algorithm includes the clothing’s main color identification

algorithm and clothing matching rules. The matching rules are used to determine the

matching between the customer and the clothing。The recommendation results are

sorted according to the matching degree.

3 Clothing Recommendation

After the knowledge bases are decided, the system needs to search the clothing

database for the clothing which meets the requirements according to the customer’s

physical characteristics.

3.1 Matching Degree Calculation

The matching degree quantifies the suitability of the garment under the user factor

through a specific score. The high total matching degree indicates that the garment

is more suitable for the user and will be recommended preferentially. The single

clothing matching degree set C{C1, C2, C3, C4} corresponds to the matching

degree of style, skin color, body type, and age respectively, within the range of [0,1].

The total matching degree of clothing formula is as follows [7]:

C C1 + C2 + C3 + C4 (1)

For example, the age matching degree is used to determine the match between a

user of a certain age and a suit for a certain age. The age matching degree is divided

into four levels, as shown in Table 1. Taking a 26-year-old user as an example, the

cloth with an age range of 25–29 is the best choice which matching degree is 1;

the cloth with an age range of 18–24 or 30–34 is the second choice which matching

A Clothing Recommendation System Based on Expert Knowledge 5

Fig. 2 Expert knowledge base

degree is 0.6 and so on, the cloth which age range is older than 40 is not recommended,

and the match degree is zero.

6 T. Yang et al.

Table 1 Age matching

degree division (26-year-old

user)

Age range that the cloth suits Matching degree

25–29 1

18–24 or 30–34 0.6

35–39 0.3

≤7 or >40 0

Fig. 3 Clothing filter process

3.2 Clothing Recommendation Process

Figure 3 shows the clothing filter process. A represents the system’s clothing

database. First, according to the filter of category factors, clothing data A1 is achieved.

A1 excludes clothing which does not meet the category. On the basis of A1, the style

is selected. The matching degree C1 of the clothing for the style is calculated. The

clothing which is unsuitable to this style is excluded. The clothing data A2 is obtained.

After preference color, skin color, body type, and age factors are selected, skin color

matching degree C2, body shape matching degree C3 and age matching degree C4

are calculated. Data with a matching degree of zero is excluded. The final clothing

data A7 is obtained which need to be sorted by the sum of C1, C2, C3, C4. The

Top-N clothing can be recommended based on A7.

3.3 Clothing Recommendation Result

Taking into account the current user’s usage habits, the system is developed based

on mobile platform. For example, user A has a warm skin and an H-shaped figure.

She is 23 years old and prefers vintage style. In the spring she needs a shirt with blue

and purple color. The results are shown in Fig. 4. The recommended clothing needs

to meet the direct filter conditions of the shirt category and blue-violet preference

color. The loose clothes can make up for the shortcomings of her thin upper body.

The main colors of both garments conform to the user style and the basic color of

the skin color.