BoF Logo

The Business of Fashion

Agenda-setting intelligence, analysis and advice for the global fashion community.

BoF and Google Partner on Artificial Intelligence Experiment

BoF and Google join forces to explore and showcase the potential of machine learning in the fashion industry.
  • BoF Team
Innovation Partner

OXFORDSHIRE, United Kingdom — Today on the VOICES stage, BoF and Google announced a partnership designed to explore and demonstrate the potential applications of artificial intelligence in fashion, and begin a dialogue between the industry and one of the global leaders in machine learning. In its first instance, the partnership has prompted a series of experiments with data sets from BoF's Fashion Week coverage, the early fruits of which were unveiled before VOICES attendees here in Oxfordshire.

Representing Google was Amit Sood, who, in 2011, founded what became the Google Cultural Institute, a non-profit arm of the company, now housed in a grand hôtel particulier in the 9th arrondissement of Paris, that has partnered with over 1,300 museums and foundations to digitise everything from the Dead Sea Scrolls to Marc Chagall’s ceiling at the Opéra Garnier, making them accessible on a platform called Google Arts & Culture. The institute is now experimenting with what machine learning can enable when applied to this catalogue, with a focus on fashion as well as art.
Sood showed members of the VOICES community some of the work his team of engineers have been doing and, alongside BoF's Imran Amed, shared the results of applying Google’s machine learning algorithms to over 70,000 runway looks from BoF's coverage of the last fours years of shows. 
"If artificial intelligence is becoming such a prominent way to solved problems, what problem could it solve for the fashion industry?" Amed asked. "I asked the Google team, 'What could artificial intelligence tell us about Fashion Week through colour?'"
The data is displayed as a visual map with clusters of colour-coded squares — "an artwork in it its own right," Sood said — which users can click on to surface runway images. They can also search by designer and material. 

"As the algorithm evolves and gets trained, you can start looking at difference things," Sood said. "None of this is human intervention. It’s like getting a baby to take more and more steps."

"It’s an experiment that we’re going to continue," Amed added. "It's really kind of amazing and would not have been possible  without a structured, organised data set."

In short, the term “artificial intelligence” or “machine learning” describes an algorithm’s ability to think, learn and adapt in complex environments where data sets can be large and messy. Google is a pioneer in artificial intelligence, which helps to power popular products like Google Maps and Google Translate, reflecting nearly two decades of machine learning research completed by the company.

We sat down with Amit Sood to hear more about Google’s partnership with BoF and the potential of machine learning in fashion.

BoF: Where are we with machine learning? What stage of development have we reached?

AS: The concept of teaching an algorithm or a network to recognise patterns and extract insights from large amounts of data is not new. Machine learning did not suddenly appear yesterday. What is increasingly happening now is that people are beginning to see actual results that stem from the machine learning insights. We are still at the pretty early stages of its development, but Google has been investing in machine learning steadily.

On the cultural side there are a couple of areas that are getting very interesting. Enabling access to the world’s leading museum catalogues is one area that could be an immediately beneficial application of machine learning, especially when it comes to extracting insights from large sets of cultural data, whether it is photography or fashion archives. Machine learning techniques can discern connections, patterns, trends and insights that would essentially be missed by the human eye or very difficult to accumulate due to scale.

BoF: Tell me more about what can machine learning can bring to fashion specifically.

AS: There is a commercial side — distribution, supply chain, improving efficiency — where I think machine learning has had and will continue to have a tremendous impact. But the part where I hope the technology will have the most lasting impact is in helping the creators — the artists, the designers and their teams — to find new forms of creative inspiration, and new ways of accessing and expressing it.

If a designer was able to very quickly analyse lace patterns from the last 500 years without having to open 500 books and spend three months researching, instead quickly pulling up a system that allows them to visually search, analyse and create a pattern based on learning from centuries ago, that is a very powerful tool. It will keep the human creativity part in tact, but it will provide creators with new ways to find inspiration.

BoF: Why did you decide to partner with The Business of Fashion?

AS: When we dipped our toe into this sector two years ago doing a project called “We Wear Culture,” which was about bringing the underlying cultural stories in the clothes we wear to a wider audience, we found fashion to be an incredibly human and creative industry, but one which has always been an early adopter of new technology and new ideas.

BoF has a very progressive approach towards technology in the sector, so it was completely natural. There was also, most importantly, an amazing opportunity to work on a data set collected by the BoF over four years of runway coverage, which we had never seen before. We thought it would be great to show very practical applications of a very strong emergent technology so that the industry would understand and get excited by it.

BoF: Describe the "experiments" you are conducting as part of the partnership. What have they revealed?

AS: When Imran [Amed] explained to us that editors and buyers spend about a month globetrotting around, watching shows and then they are expected to compile and analyse all of this "data" and report on what the biggest trends for the season will be in terms of colour, shape, materials, etcetera, almost in real-time.

This got us thinking, what if we apply machine learning to this data set and see what happens? Would the machine find useful insights for the editors? Could designers, creatives and students use computers to analyse large sets of fashion data and use that for inspiration and deeper understanding? And could techniques like this tell us more about the business — about trends, consumption habits and future predictions perhaps?

You have the technology and you have the content or data, and then I think the biggest factor that really drives the success of the experiments is bringing in artists and practitioners — what we like to call creative coders, artists in their own right, not just engineers, but engineers that see their work from a creative perspective.

We need to keep experimenting; the technology is there to help, in ways that we may not be able to understand immediately. It is important that the sector stays with technology as it evolves rapidly. This is something that we care about a lot. There is a potential risk that the technology will keep moving on at the crazy pace that technology does and certain businesses in the cultural and creative sector, which have their own way of working and their own pace, may struggle to keep up. It is important to bring them all along.

BoF: What excites you the most about the potential of machine learning in the long term?

AS: We are at risk of forgetting a lot of our past, a lot of our history, which is stored in the equivalent of locked lockers that are not accessible. The data exists, the big cultural institutions all have huge archives, fashion houses have huge archives, there are documents and documents, sketches and almost infinite ideas. So much available data, but data that is rarely prioritised as the funding for cultural industries is simply not top of mind today. This is where the technology can play a role.

Machine learning can help organise that data, analyse it, and then institutions can choose what to make accessible. And I think that is a real benefit for the public. This is step one, but you are going to see rapid progression in the technology — faster than anything else we have seen before because these algorithms train themselves with data sets very quickly.

Google is the Innovation Partner of VOICES 2017, an invitation-only event bringing together the trailblazers of the fashion industry, uniting them with the entrepreneurs and inspiring people who are shaping the wider world. To learn more about the gathering, visit our VOICES website.

QIC Global Real Estate is the Principal Partner of VOICES 2017.

Related Articles:

© 2021 The Business of Fashion. All rights reserved. For more information read our Terms & Conditions

The Business of Fashion

Agenda-setting intelligence, analysis and advice for the global fashion community.
The State of Fashion: Technology
© 2022 The Business of Fashion. All rights reserved. For more information read our Terms & Conditions and Privacy policy.
The State of Fashion: Technology