How it Works: AI Fashion Styling

 

A question that I often get asked when I attend events or explain Intelistyle for the first time to people is “Is this the real thing. Is it real AI?”. Which used to surprise me but then I realised that there are so many companies out there that claim to do Machine Learning, but what they really do is package up a couple of Amazon, Google or Microsoft cloud ML APIs, combine them with a freely available dataset, and voila, here’s a service that they can sell to customers.

 

So I decided to do a write up of how our technology works to help our customers understand how this really works.

 

We constantly crawl the web, very much like google’s search engine does. Instead of indexing generic information though, we focus on fashion data. We have particular data sources that we prefer, like fashion magazines, social networking websites, retail websites and blogs. That process allows us to collect thousands of outfits put together by human stylists. We use images and text to get the most complete and accurate information.

 

Now as you can probably imagine most of the web’s images are quite noisy. How do you extract the individual garments that are included in an outfit with varied backgrounds, different poses and models. The approach we took, was to create a bounding box model that can create bounding boxes for each garment. Using that approach we were able to create a unique dataset of millions of outfits that we could use for training our AI styling model.

That dataset is constantly updated and quality controlled by our team. That allows us to keep up with the latest trends across different regions. We have clients in China, Europe and the Middle East and as you can imagine the trends in each of these regions are very different. What is considered fashionable in one region, isn’t necessarily in another.

 

Our Machine Learning team uses the latest academic research to craft a proprietary, bespoke set of AI models that analyse images and text. Each garment in our database is described using a 128-dimension “signature” or embedding. You can think of this as a very similar process to what Shazaam does for music tracks. Each of these signatures describes the important characteristics of each garment and leaves out the noise.

 

However to create styling intelligence that can perform as well as actual stylists a good dataset and embeddings was not enough. While talking to our clients we realised that there are fashion rules that can make or break an outfit. For example, an off the shoulder top with puff sleeves should not be styled with a skinny-fit blazer. Our model could not predict these rules as good as humans yet. 

 

The solution to that was to train another model that can detect rich attributes, such as it’s fabric, cut, style, colour and other unique characteristics and categories for each garment of our dataset. 

 

We also work with stylists with experience in brands such as M&S, Topshop and Vogue to create a unique set of ‘guidelines’ for our model to give preference to specific attributes when creating an outfit. And off course because no two regions are the same, we can customise those guidelines to particular trends. For example in the Middle East shorter hemlines always pair with a longer overcoat and in Asia slip dresses should be layered over shirts.

 

The result? A proprietary set of AI and data, that outperforms all published academic research and delivers outstanding styling quality, trusted by the world’s top luxury brands such as D&G, MaxMara and Lane Crawford. 

 

We even tested it against real stylists and fashion influencers at London Fashion Week. As Forbes reported, 70% of respondents unwittingly chose the looks created by our model. 

 

Blockchain, A.I. and Other Game-Changing Fashion Tech in 2019

The world’s most significant and profitable industries are facing massive changes thanks to advances in technology. More specifically, blockchain, artificial intelligence (A.I.), new types of financial transactions and a few other big leaps in tech are responsible for these ongoing changes to how many industries do business. Looking at the fashion industry in particular, these are some of the ways in which we’d expect to see specific changes come about.

Blockchain Will Fight Counterfeit Products & Streamline The Fashion Supply Chain

The distributed ledger technology called blockchain is mostly known for being the foundational tech behind cryptocurrency. More generally, a blockchain is basically a string of chronologically arranged code. Each new block of code on the chain requires every computer with access to the blockchain to approve the addition of that data via a shared cryptographic solution, which translates to peer-verified security in every data transfer, addition or modification. It all happens nearly instantaneously, without any of the red tape commonly attached to security protocols. And this same tech can be used to protect products and ensure quality in the fashion industry.

Each individual fashion item can be tagged with labels attached to blockchains, which allows everyone in the supply chain to verify its origin, ownership and every time it has changed hands. The LuxTag Project took to Medium to detail how some designers are already taking advantage of this potentially revolutionary product integrity solution. Back in 2017, Londoner Martine Jarlgaard produced the first smart label-tagged garments. Scanning the tags gave users time-stamped info on everything about the garments – from raw material acquisition and factory information down to how the finished products were packaged and delivered. Similarly, fashion retailer Babyghost used near-field communication (NFC) chips to tag its 2017 summer and spring collection. This allowed customers to use the NFC tags to find out everything they wanted and needed to know about every Babyghost product.

Apart from ensuring product integrity, this can also push fashion labels to be more honest about where they get their raw materials, how they conduct labor practices and everything else customers in 2019 (and beyond) will be concerned about. Essentially, information sharing and supply chain transparency are about to become realities in the global fashion industry. This is a potentially huge boon to the fashion supply chain, the fight against counterfeit goods, and corporate social responsibility.

Competitive Payment Options Will Dictate Retail Preferences

In recent years, payment platforms such as Paypal, Amazon Pay, Payoneer, Venmo and Dwolla have played increasingly large roles not just in online retail, but in various other spaces such as vacation rentals, gaming and service industries. The reason is simple: these payment options offer the convenience and security that some more traditional methods lack. By and large, newer, more innovative payment platforms are now making things even easier for both customers and product and service providers.

For instance, the service called Paysafe Pay Later is potentially revolutionary in allowing customers to delay payment until after their ordered products have been shipped, all in a way that doesn’t impact the company’s cashflow. A New Zealand-based gaming company details how online casinos now use the Paysafecard, which is similar but slightly faster to use than VISA and other debit/credit cards. Independent of bank, card or other personal financial information, Paysafe instead relies on a single, 16-digit pin to credit money to a customer’s account and verify transactions. It’s currently seeing heavy use in gaming and retail – two of the biggest revenue sources on the web. And it’s just one example of a new payment service that could prove to be a deciding factor in how customers choose which fashion retailers to shop from in the future.

Nobal’s smart mirror solution for fitting rooms

Artificial Intelligence Will Change Everything

There are several reasons why Intelistyle cannot help but cover how A.I. will change the face of retail. For one thing, A.I. algorithms are responsible for well-informed and relevant product suggestions for online retail customers. For another, A.I. can be used to predict trends and product demand, allowing retailers to be better at managing inventory and catering to customers’ needs.

Though they may seem simple, even those small perks can help retailers avoid serious problems – such as, for example, the massive surplus Under Armour faced last year when it overestimated product demand and wound up $1.3 billion the hole! And even this is only scratching the surface of the benefits A.I. can provide to retailers.

Examples could go on and on, but even the relatively brief write-ups above provide a picture of how a small handful of tech innovations can and will revolutionize the fashion retail business.

How Artificial Intelligence Is Set To Change The Next Decade of Fashion (with Paul Kruszewski)

 

An exclusive interview with Paul Kruszewski, AI technologist, serial entrepreneur, founder and CEO of WRNCH – a leading AI computer vision software engineering company based in Montreal.

The “machines” are coming! As the farfetched robot apocalypse of Sci-Fi films becomes an impending reality of our daily life, the reaction of the fashion industry remains mixed. When a technology is projected to revolutionise the industry, the possibility of feasting on the early adopter’s advantage captures the most sceptical hearts… on the other other hand, scepticism takes the reins and demands to know:

Will AI be the end of human creativity and jobs in fashion?

While the answer is more complex than a single word, it is still a solid: No.
The good, the bad and the ugly speculations surrounding AI should all be taken with a grain of salt.

Darwin’s foresight that it’s neither the strongest nor the smartest but the most adaptable to change that will survive the test of time rings true. The fashion industry is dipping its toes (or legs as of 2018) in the waters of artificial intelligence.

We sat down with a man who took a deep dive into these novel waters 20 years ago: Paul Kruszewski, AI technologist and serial entrepreneur. During our exclusive interview on his personal journey to success, he shared valuable tips for fashion companies looking to incorporate AI into their business. Come and take a look at the past, present, and future of AI and its implications for the fashion industry.

 

 

From Farming To Artificial Intelligence

Paul’s first encounter with technology dates back to late 70s, when he was raising pigs for pocket money as a 12-year-old kid at his family farm in Alberta. “I really wanted a computer, that was the cool thing” he recalls. After selling his pigs for $250 and combining it with his brother’s $250 input, he was still only halfway there. “I go to my dad and said, programming is the future. We need to buy a computer. You match us up to $1,000”. $500 later, he was writing his first program on a Radio Shack TRS-80.

 

Radio Shack TRS-80 is one of the first desktop microcomputers launched in 1977

 

Fast forward to 1998, having completed his bachelor’s in computer science at University of Alberta, his MA and Ph.D. at McGill University, Paul was recruited by a company called My Virtual Model: “They said, e-commerce is going to change everything. We’re going to create your body on the internet and we will put clothes on you and we will charge”. With just a powerpoint and a $30 million capital, he took the team from 1 person to 60 in 9 months, built a chip while the founders sold the software. Looking back at the business, he maintains the vision was “conceptually great” but “the promise of trying clothes on the internet and making a buying decision from that… It was 20 years too early. The technology wasn’t there!”.

 

See how Nobal is using iMirrors to enhance the in-store customer experience in our interview with Thomas

 

An Industry Right on The Cusp of Change

When asked where he sees the technology today in means of enabling that vision, he says “We’re right on the cusp”. Paul goes on to predict that in 5 years it will be pretty standard for people to stand in front of their TV and try on basic clothing. “I think AI is going to transform everything… the supply chain, design. The design process of clothing will be semi-automated in 10 years. On the other side of the spectrum, AI is going to completely change materials and 3D printing”.

 

Electronic textiles known as “Smart Fabrics” enable digital components to be embedded in them for a variety of benefits from customised fit to weather adaptability

 

Business of Fashion’s collaborative report with McKinsey & Company confirms Paul’s prediction, stating that 20-30% of current fashion jobs will become automated. But, don’t brace yourself for such a disaster scenario just yet. Rather than replacing humans, AI will be supplementing existing jobs and will be creating brand new ones. The bottom line is that fashion still does and will persist to need the human touch. Prominent fashion schools of the world like The Fashion Institute of Technology in New York and The London College of Fashion are already incorporating AI-led skills training into their degree programs to raise the next generation of industry leaders.

“…who controls the customer experience? Will it be the clothing manufacturer, the retailer or the person behind?”

Speaking of leaders, back to Paul Kruszewski, who was next recruited as the CTO of a video games company. “But I realised that I’m a technologist and an entrepreneur. In 2000, I started my first company in AI”. Shortly after, he was approached and persuaded by Rick McKenzie, a professor, and military researcher, to collaborate on military simulations. It didn’t take long to start getting results…

 

The Technological Power Play

When once he was struggling to secure a place in the market, he was now getting calls from companies and corporations including the US Marine Corps. But what is the secret to his commercial success? “No matter what you think is the right call at the time, you have to listen to the market” he reveals, “There will always be that interesting tension of who controls the customer experience. Will it be the clothing manufacturer, the retailer or the person behind?”. While we will have to wait to find out the definitive answer to that question, he expects to be interesting to watch as completely different possibilities for fashion unfold over the coming decade.

 

The Vital Questions To Ask Your Business

As with all virgin territory, adapting AI into your business comes with the innate possibility of mistakes along the way. “All of a sudden everyone is an AI-powered company. Why?” he inquires and then proceeds to highlight that AI is a tool to enhance your existing product and not a product in itself. In this uncharted land, where there isn’t yet enough history to go by and map out a fine print to success, Paul encourages companies to go back to 3 basic questions for a seamless integration process:

1- What are you trying to solve?
2- What is your pain as an organisation?
3- What are you doing that you want to get better at?

And find out where you can be unique!

 

A Curated Dataset Is The Key To Success

A lot of traditional fashion and retail organisations do not have the expertise to work with their data. In such instances, Paul recommends collaborating with an external AI expert who will curate their data for optimal results. Which brings us to one of the most immediate challenges in AI, collecting a good data set: “There are widely available datasets but your competitors also have access to them” he says. Bespoke, customised data is the way forward for finding and capturing the unique competitive advantage that your business needs to differentiate and come into prominence. Once again, delegating the task to AI experts specialised in your industry, who can help you build your own data sets is key: “Some organisations initially freak out because they have never heard of that. And that’s fine!”

 

Companies are encouraged to work with an AI expert on curating their own datasets

 

The Evolution of AI in Fashion

Talking about the evolution of AI as it relates to fashion, Paul reports that we have come a long way from the early simulators that were essentially just a pose with no interactivity. “We generate synthetic data: virtual humans, virtual clothing. We dress them, put them in virtual environments and we take virtual pictures of them”.

See the full interview here.

Bitnami