See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/345895110

AI FOR FASHION

Conference Paper · September 2020

DOI: 10.6084/m9.figshare.14540733

CITATIONS READS

0 795

1 author:

Edit Dr. Csanák

Óbudai Egyetem
31 PUBLICATIONS   43 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Sustainability in fashion View project

Denim manufacturing View project

All content following this page was uploaded by Edit Dr. Csanák on 15 November 2020.

The user has requested enhancement of the downloaded file.

 13th INTERNATIONAL SCIENTIFIC-PROFESSIONAL SYMPOSIUM TEXTILE SCIENCE AND ECONOMY
18th September 2020, ZAGREB, CROATIA

AI FOR FASHION

Edit CSANÁK

Abstract: Technology and Artificial Intelligence have a significant impact on all aspects of Fashion, from
designing to production and consumption. Fashion has always been a forward-looking phenomenon, willing to
adopt new technologies as they emerge. Artificial Intelligence is no exception as it moves as fast as
Fashion. The AI has been used in analyzing fashion trends and consumer needs for over a decade. Its impact
on style and the enhancement of phenomena as Fast Fashion, is indisputable. The clothing industry 4.0
digitalization is increasingly relying on the use of advanced technology, moving towards its 5.0 version, the
use of Artificial Intelligence in the broader scope of activities. However, the study of the short career of Artificial
Intelligence in Fashion clearly shows that AI has a significant impact on this phenomenon and the related
industry, a considerable player of the global economy. The current situation raises numerous questions about
one of the primary cultural phenomena. Some fashion professionals question the current unpredictable social,
cultural, economic environment evolving the future of fashion. Along with the issue of sustainability, several
are addressing the issues raised by the application of artificial intelligence. Thus, the examination of the
supposed directions, with the prediction of their wither impact, is essential for the development of creative
industries. This article attempts to review the current use of AI in the fashion industry and fashion market. This
article is part of a study conducted by the Doctoral School of Security Sciences, the Artificial Intelligence
Workshop, and the Product Design Institute of Sándor Rejtő Faculty of Light Industry and Environmental
Engineering of the Óbuda University Budapest, Hungary.
Keywords: fashion design, digitalization in Fashion, AI, Artificial Intelligence

1. Introduction
Fashion, which motor is the continuous change and progress, faces new challenges day-by-day. The fashion
industry today is one of the leading economies in the World, with an estimated value of up to 3,000 billion US
dollars [1,2]. Garment manufacturing, one of the oldest human activities, has come down through the centuries
with continuously adapting to technology and society improvements. Striving to keep up with the fast-changing
consumer needs, the fashion industry today is rapidly adopting the post-modern, Industry 4.0 and recently 5.0
garment production technologies, along with all the latest digital achievements. The fulfillment of economic,
social, and cultural sustainability in a holistic sense is of unquestionable interest to the World today. All the
real-time efforts direct towards eliminating the polluting activities and production processes of several well-
known harmful industries, such fashion industry is. The number of initiatives and actions taken to reduce the
environmental activities of the fashion industry has increased vividly over the last few years. Even the tangible
results the players involved still do not seem to be pulling the wagon in one direction. Over the last decade,
significant digital innovations have been applied not only in sales and retail but also in digitalization of the entire
phase of the clothing supply chain. These are extended to the technologies used in designing. Bodies of
fashion and technology experts examine how to integrate the latest technological innovations to fit the demand
of revitalization of the fashion industry, extending it to the involvement of the AI. However, when technology
and Artificial Intelligence are applied in an uncontrolled way, modern technologies only accelerate the already
overburdened economic processes. They may contribute disrupting the sensitive economic system of the
fashion industry, and calling into question its cultural, artistic, and aesthetic value.

2. Basics of Artificial Intelligence

Artificial Intelligence (AI), often called machine intelligence, is the kind of Intelligence that machines show. The
general purpose of Artificial Intelligence, an academic discipline that was founded in computer science in 1956,
is to create technology that enables computers and machines to operate intelligently. According to the current
state-of-art, AI is considered any device that perceives its environment and takes actions that maximize its
chance of successfully achieving its goals. In short, therefore, AI is a machine that can learn and think, and
can perform specific tasks independently. The traditional goals of AI research include reasoning, planning,
learning, knowledge representation, language processing, perception, and the ability to move and manipulate
objects. Further areas of research are Social Intelligence and General Intelligence. A system that recognizes
its environment and takes actions that maximize its chances of success is called an intelligent agent.
Programs that solve specific problems are the most simple intelligent agents. More complex agents can imitate

117
 13th INTERNATIONAL SCIENTIFIC-PROFESSIONAL SYMPOSIUM TEXTILE SCIENCE AND ECONOMY
18th September 2020, ZAGREB, CROATIA

humans, copy organizations of human beings (for instance, a company), or copy their activities. Intelligence
subdivides the general problem into sub-problems. They consist of features or capabilities that researchers
expect to see an intelligent system appear. Fields like planning, learning, perception, motion, or manipulation
of objects are just some of the featured areas of AI research. Knowledge representation and reasoning in
the field of Artificial Intelligence (AI) are committed to presenting information in a form that a computer system
can use to solve complex tasks. Problems of reasoning and problem-solving were solved already during
early development; the AI research was able to model the process of reasoning and problem solving (Allen
Newell and Herbert A. Simon, 1959).

AI research has developed methods for dealing with uncertain or incomplete information, using the concepts
of probability and economics. However, the first algorithms proved inadequate to solve big argument problems
since they faced a "combinatorial explosion": when the problems grew more significant, the machines became
exponentially slower. Most of their problems are solved with quick, intuitive judgments.

The basis of the programming language is conditional statements, conditional expressions, and conditional
constructs. The if-then construct (sometimes called if–then–else ) is common across many programming
languages. It is the core boolean condition that a computer can solve. It bases on the intrinsic values that
something is True or False . In contrast, the combination of this expression, the If and Then surrounding
it, and the subsequent consequences, form a conditional statement that has an inner meaning (for example,
expresses a coherent logic rule) but has no intrinsic value.

2.1. Algorithms

The processes which enable computers and machines to operate intelligently are algorithms. Different
processes assigned to the CPU (Central Processing Unit) are based on particular scheduling algorithms. The
algorithm is a sequence of steps for a computer program to complete a task. The writing of algorithms is a
science within computer science: creating unique algorithms and knowing when and how to apply them are
the base of the computer programs which make our everyday life well-processed today. By the use of
smartphone applications, we daily use dozens of them while running the videos (audio and video compression
algorithms), or finding the best route to our office (route finding algorithms). Optimization and process
scheduling algorithms arrange the traffic lights during our travel. These algorithms may be non-preemptive or
preemptive.

3. Artificial Intelligence for Fashion in the era of Big Data

By Nassim Nicolas Taleb (in a complex meaning), Intelligence consists of ignoring irrelevant things. Artificial
Intelligence incorporates a set of methods that are very suitable in the field of Fashion. AI is capable of handling
the "3V" of big data, namely speed, variation, quantity with uncertainty, volatility, complexity in the fashion
industry, and related markets. However, AI techniques are already in use in all sectors of the fashion industry,
the diversity of currently available methods, models, applications, and data types makes implementing these
techniques difficult. It scares some fashion companies. The consumer movements, such as sustainable
fashion movement, make fashion consumers in their shopping actions to be more increasingly driven by a
sense of consciousness rather than greed. Nevertheless, it still can be stated people have never consumed
as many clothes as today, and along with it, the way how they do it changed.

3.1. Internet in action: Data mining and fashion sales predicting methods based on AI

The information has become probably the most valuable element for society and industrial progress today.
The big data environment has radically transformed our daily life and the economic and business World. For
the fashion industry, the Big Data Era, supported by applications of the Internet, is very challenging and poses
a considerable scale of opportunities. Many types of data can be analyzed in the context of Fashion: point-of-
sale (POS) data, geographic information systems (GIS) data, social media data, virtual 3D data, sensory data,
textile physical data. The management of the profitable use of these data requires advanced techniques. Data
mining is an interdisciplinary subfield of computer science and statistics, the process of discovering patterns
in large data sets involving methods at the intersection of machine learning, statistics, and database systems.
The overall goal is to extract information (with intelligent methods) from a data set and transform the information
into a comprehensible structure for further use. Data mining is the analysis step of the "knowledge discovery
in databases" process or KDD. Before data mining algorithms can be used, a target data set must be
assembled. Application of Knowledge Discovery in Databases (KDD) – a form of data mining, roots back to
the 1990s, when the term data mining appeared first in the database community; these were the years 2005-
2006. [3,4].

118
 13th INTERNATIONAL SCIENTIFIC-PROFESSIONAL SYMPOSIUM TEXTILE SCIENCE AND ECONOMY
18th September 2020, ZAGREB, CROATIA

The success of a collection is a crucial issue for the survival of a fashion brand. To predict the success of
fashion products is difficult, as the ever-changing landscape of Fashion shows short life cycles and high
volatility of the consumer trends. The foreseeing abilities of the skilled predictors refined in the digital era; the
traditional techniques developed, more recently involving the Artificial Intelligence in the Big Data analysis.
Several AI methods and AI-based hybrid methods have proven effective in forecasting fashion sales
performance. [5] [6] With the increasing demand to be able to follow consumers' demands, the quality and type
of information changed. Before companies expected to be told, through valuable reports and trend-books,
what to design for their customers in detail, today, designers prefer to rely on comprehensive lifestyle, socio-
cultural, and economic trend-analysis building their brand identity alone. Today, fashion industry workers and
trend forecasters rely on the Internet to retrieve information; they spend a significant volume of time tracking
what is being looked at online and who is doing the looking. New technologies are helping retailers to manage
inventory using AI-powered tools to gauge demand.

3.1.1. Virtual style assistants and fitting applications

Forecasting is but just one example of using Artificial Intelligence in the fashion industry. According to an article
recently published by the Forbes Magazine: "…despite the established nature of the fashion industry, AI is
fundamentally transforming the industry from the way that fashion companies manufacture their products to
the way they are marketed and sold. AI technologies are transforming the fashion industry in every element of
its value chains such as designing, manufacturing, logistics, marketing, and sales." change is coming (whether
we like it or not)… Article of the WWD magazine highlights that the world-leading fashion industry sees AI
empowering designers, brands, and retailers to make better products and create more compelling shopping
experiences. [7] AI has reached the fashion industry making the shopping experience of the customers ever
comfortable. For example, a virtual DressingRoom app of the clothing giant Gap, introducer in 2018, created
to help clients to try on the desired clothing virtually. (Fig 1) [8]

Figure 1: GAP's Virtual DressingRoom visualizes the outfit on one of five body types

Another application example is the Nike Fit, a scanning app that collects data on 13 points on a person's foot
to measure the full shape of a user's feet within a matter of seconds, fitting each Nike shoe style using a
combination of computer vision, data science, machine learning, Artificial Intelligence, and recommendation
algorithms. (Fig 2) [9]

Figure 2: Nike Fit foot scanning application

119
 13th INTERNATIONAL SCIENTIFIC-PROFESSIONAL SYMPOSIUM TEXTILE SCIENCE AND ECONOMY
18th September 2020, ZAGREB, CROATIA

3.2. AI as a fashion designer

Fashion experts are skeptical about if an industry built on creativity can be ever computerized completely.
Fashion, as a materialization of human fantasy concerning the human body, shaping of the silhouette of the
outfit, and creation of custom beauty, is eternally an art, an activity addressed to designers and artists. If AI,
as augmented Intelligence, can help the human thinking process to focus on higher-value decision-making,
arises a logical question: Can AI become a fashion designer? The answer is, unfortunately: Yes. The San
Francisco based research center of the giant merchandising Amazon was launched the year 2004. Lab126
has developed an algorithm able to recognize particular fashion styles tracing on images; later, the tool can
generate similar style new items. A simple, however competitive AI fashion designer (Fig 3).

Figure 3: Amazon's fashion algorithm that can design clothing by analyzing a bunch of images to copy the style and then
apply them to new items

AI-enabled improvements are already found in the fashion sector; several forward-thinking retailers are already
using social networks (Instagram, Pinterest) to track the latest fashion trends and respond quickly. Processes
of the leading companies are powered with smart technologies, enabling higher speed, lower cost, and
improved flexibility at each stage of the supply chain, and it has further capacities in production automation
and delivery of the goods. It has further capabilities in areas including forecasting, analyzing of new trends
choosing and making right and sustainable fabric and color combinations, designing the desired cut with zero
waste, organization of the production process in the most flexible, and sustainable way (Fig 3).

4. Highlights of AI in relevance to fashion designing

When discussing prospects of application of Artificial Intelligence in Fashion Designing, firstly, we must
highlight: Design is a form of reasoning-argumentation of a solution, a kind of a proposal. The arguments can
also be typed according to what logical way we reach the conclusion we formulated in the theorem. Fields and
features that receive the most attention in AI research are processes of planning, reasoning, and problem-
solving, which also constitute the most complex problems in the design. Design is a form of problem-solving.
In a psychological approach, we call a problem any question or task for which we cannot find the exact or
immediate answer or the solution. The types of problems and critical phases of the thought process: 1. Findings
(diagnosis), 2. Classification (analysis, sorting), 3. Design (goals, solutions). Figure 4, with a range of
sweatshirts, illustrates the problem-solving issue and process. (Fig 4)

Figure 4: Range of sweatshirts (left to right): 1. Plain, gray round-neck long sleeve, 2. Decorated, white round-neck long
sleeve, 3. Plain, yellow long sleeve hoodie, 4. Decorated, dark gray long sleeve hoodie with zipper

120
 13th INTERNATIONAL SCIENTIFIC-PROFESSIONAL SYMPOSIUM TEXTILE SCIENCE AND ECONOMY
18th September 2020, ZAGREB, CROATIA

According to the general methodology, the macro-phases of the problem-solving process are the following:
1. Finding - Diagnosis:
 What: is the problem that needs to be solved? (See: To create a sweatshirt.)
 Who: is the general audience who needs to solve the problem? (E.g., for men, women, or children.)
2. Classification - Analysis, sorting, and systematization:
 Vision: What is it? (See, for example, to create a Spring-Summer or Fall-Winter collection.)
 Strategy: How will we do this? (E.g. creating a custom product range for a customer group.)
 Target: Who is the specific target group? (Subdivision; e.g., men, young woman, teenagers, etc.)
3. Design - Defining outcome goals, suggestions for solution:
 Goals: What are do we set? (See: To create sweatshirts, long sleeves, round necks, etc.)
 Suggestions: What exactly are we going to achieve? (For example plain, or decorated, striped, etc.)
 Features: What features does this manifest have? (Round neck, hoodie with zipper, white, etc.)

In the case of a basic sweatshirt (Fig. 5), a problem seems relatively easy to be solved. Arising trend-
related features are "easy" to classify stylistically, harder in aesthetical approach! See Figure 5.

Figure 5: Example of decorated long sleeve, round neck sweatshirt

4.1 Knowledge representation: Highlights of fashion designing with machine

A machine a computer program able to reason, and which uses a knowledge base to solve complex problems
is called a knowledge-based system (KBS). Technical, scientific, and social aspects involved in building,
maintaining, and using knowledge-based systems refers to knowledge engineering (KE). Artificial
intelligence systems imitate complex human decision making. It bases on three types of reasoning: induction,
deduction, and abduction. The most challenging problems of knowledge representation are [3]:

1. Default reasoning and the qualification problem, i.e., many of the things people know take the form of
"working assumptions." In design, it refers to a matter of the use of subjective constancies.

2. The breadth of commonsense knowledge (i.e., knowledge necessary to run an average research project)
builds on examination of a vast amount of information. The average person knows a lot of atomic facts.
Humanity has accumulated an almost unmanageable amount of clothing-type information during civilization.
Organizing this is both a formal and a stylish challenge. See it illustrated with the example collar, which is only
one of the (possible) details on the garment (Fig 6).

Figure 6: Collars (left to right) Tudor ruff (Portrait by Michiel Jansz van Miereveldt of Elizabeth Queen of Bohemia 1623)
Edwardian collars (around 1911), Historic details from 1911 and 1912, and 1920s Men's Shirts and Collars

121
 13th INTERNATIONAL SCIENTIFIC-PROFESSIONAL SYMPOSIUM TEXTILE SCIENCE AND ECONOMY
18th September 2020, ZAGREB, CROATIA

Symbolic artificial intelligence is the expression of a collection of all the methods used in the research of
artificial intelligence, based on a high-level (symbolic) (human-readable) representation of problems, logic, and
research. 3. The subsymbolic form of some commonsense knowledge, i.e., what people know, is not a
fact or statement that they can verbally express. Many of the decessions/expressions are made driven
unconscious and sub-symbolic intuitions or tendencies in the human brain. Such knowledge informs, supports,
and creates the context of symbolic, conscious knowledge. In terms of design, it features:
 Designing often bases on a hypothesis (abduction), a "violent," a formal conclusion, a kind of "best
explanation." The creative mind can reach a proper conclusion without logic, ignoring the rules. The
design process is the use of simple logic in a system of visual elements and principles. Any reasonable
person will identify similar variables and interactions.
 Elements of abductive design will make the final result will be different. These are 1. The
designer, 2. The design process (itself), 3. The product to be designed. These elements may vary and
are always different due to the approach of the designer, resulting in that the final product will always
be different.
 The process of creative cognition is a somewhat unconscious process, characterized by a high
degree of adaptation. The creative process is a mental process based on the interaction between
analysis and imagination, characterized by consciousness, passion, and commitment.
 Creativity as the implementation of "unexpected coexistence." To create a new product, we need
to have some knowledge, but we must also forget this knowledge to realize the novel, unexpected
relationships between things.

5. Challenges on the application of AI in the future fashion design

In the age of digitization, AI and machine learning (ML) based technologies offer manufacturers automated
solutions in the fashion industry. Artificial intelligence is currently used in many areas of fashion, and
professionals already see the best possible opportunities to exploit it further. AI-compatible machines and
robots easily sew fabrics perfectly, while also being able to detect fabric defects and provide quality assurance
to ensure that actual design shades match the new colors. Some retail giants, such as Amazon and Walmart,
already have their clothing brands and use machine learning systems that identify the venue and, in the not-
too-distant future, create fashion trends that customers would rather buy. The AI can detect new trends with
reduced (human) forecast error. AI can design to make endless combinations of the variables (see illustrated
with Fig. 4, 5, and 6). Changing something will change everything; the many possible connections between
design and performance variables result in different performance.

Why is it all so challenging for the future of fashion design? Since the possibility of taste reduction
destroying the culture. Application of the intelligent machine to essentially human activities bases todays on
fast teaching of existing patterns. Fashion design bases not only on random variations, mixture or custom
adaptation of these patterns. If art and design are forms of "creating beauty and order in function" (Viktor
Papanek), then fashion is continuous redefinition of the function and clothing aesthetics. The Muze project is
a pilot collaboration between the European e-commerce site Zalando and Google. It was one of the first
“machine learning” projects involved in fashion design. Muze does not assemble products; it only offers ten
different basic model versions in different colors and prints, based on “inspirations” given by users (Fig. 7).

Figure 7: Designs made by algorithm Muse

122
 13th INTERNATIONAL SCIENTIFIC-PROFESSIONAL SYMPOSIUM TEXTILE SCIENCE AND ECONOMY
18th September 2020, ZAGREB, CROATIA

Algorithm-Driven Design is the new know-how in design; it is about how Artificial Intelligence is changing the
design [11].

6. Conclusion
Consumption is rising day by day in the super-powered digital era, more recently, including the involvement of
artificial Intelligence. Increased application of Artificial Intelligence in Fashion raises new questions in the
aspect of the quality of design and its cultural aspects. Design, which is profoundly a humanistic endeavor,
can lose its essential value, which refers to Humanity and Art. The widespread application of digitalization
opened endless technological opportunities, supporting retail in the faster speed of the goods to market, what
are bad news for Sustainability in Fashion. This article attempted to summarize a handle of facts, which
constitute the basis of a recent, prospective collaborative research.

References

[1] O'Connell L.: Size of the global apparel market in 2015 and 2017, Available from
https://www.statista.com/statistics/279735/global-apparel-market-size-by-region/ Accessed: 12. Nov.
2019.
[2] FashionUnited, "Global fashion industry statistics - International apparel," [Online]. Available from:
https://fashionunited.com/global-fashion-industry-statistics/#tab2 Accessed: 8. Nov. 2019.
[3] Russell S. J.; Norvig P.: Artificial Intelligence: A Modern Approach, Vol.1. Pearson Education (2010)
[4] Bertola P.; Teunissen J.: Fashion 4.0. Innovating the fashion industry through digital transformation,
Research Journal of Textile and Apparel, Vol. 22, no. 4. pp. 352-369., 2018.
[5] S. Ren, P. Chi-Leung Hui, and T.-m. Jason Choi: AI-Based Fashion Sales Forecasting Methods in Big
Data Era, in Artificial Intelligence for Fashion Industry in the Big Data Era, Singapore, Springer, 2018, pp.
9-26.
[6] R. Schmelzer: The Fashion Industry Is Getting More Intelligent With AI, Available from
https://www.forbes.com/sites/cognitiveworld/2019/07/16/the-fashion-industry-is-getting-more-intelligent-
with-ai/ Accessed: 12. Nov. 2019.
[7] J. Standish J.; Ganapathy V.: Think Tank: How AI Can Power the Future of Fashion, Available from
https://wwd.com/business-news/business-features/jill-standish-think-tank-1202941433/ Accessed: 14
April. 2020.
[8] Sheldon P.: Augmented reality in retail: Virtually try before you buy Available from
https://www.technologyreview.com/s/614616/augmented-reality-in-retail-virtual-try-before-you-buy/
Accessed: 14. January 2020
[9] Nike.com: How Nike Fit Works, Available from https://news.nike.com/news/nike-fit-digital-foot-
measurement-tool Accessed: 27. May 2020
[10] Mutchler A.: Amazon Takes The Fashion Angle with New Echo Look, Available from
https://voicebot.ai/2017/04/26/amazon-takes-fashion-angle-new-echo-look/ Accessed: 20 April. 2020.
[11] Algorithm-Driven Design https://algorithms.design/ Accessed: 02. August 2020

Author:

Dr. Edit CSANÁK

Product Design Institute, Rejtő Sándor Faculty of Light Industry and Environmental Engineering, Óbuda University
Doberdó út 6, 1034 Budapest, Hungary
Phone: +(36) (1) 666 5930 Mobile: +(36) (30) 667 3639 E-mail: csanak.edit@rkk.uni-obuda.hu

123

View publication stats