The article first appeared on the Forbes.
Between the notion of clothing for clothing's sake and seasonal fashion exists style. A subjective and at times anxiety-inducing concept, style defines what your outfit and how it was ‘put together' says about you. Whether you prefer to put your own outfits together or be told what ‘looks good/is right for you' by a personal stylist, style is a language all of its own.
Artificial Intelligence is emerging as a powerful tool in fashion retail to shape personalized product suggestions and refine the way customers are targeted to drive purchasing. ‘Style' has largely sat outside of this data remit, perhaps hampered by attitudes towards AI that place it as a tool for emotionless decision-making, rather than a tool that can decode something as ephemeral and nuanced as style. Think again. London-based startup Intelistyle took to the streets during London Fashion Week with their AI-driven style app that lets their algorithm do the styling. But how would the fashion industry respond to ‘computer-generated looks'?
The Intelistyle team conducted a modest study by collecting data from 27 fashion week goers ranging from fashion stylists to bloggers and influencers. The subjects were presented with pairs of outfits and asked to choose the one they ‘preferred the style of'. 70% of respondents unwittingly chose the AI-derived looks, which had been pitted against an outfit pulled from an Instagram influencer. But how are the AI looks generated?
Intelistyle's approach is to "crawl the web for fashion photography examining thousands of outfits put together by stylists, influencers, designers and retailers." Using deep learning, they "extract the essence of style." How does this differ from other AI-driven style apps? Typically, computer vision techniques that focus on pattern recognition are used to find visually similar clothes. However, recognizing that style is a lot more nuanced than finding similar patterns or colors, Intelistyle has deepened their analysis to determine what can make seemingly different or ‘clashing' patterns or fabrics combine together to build a beautiful outfit. They have also devised how two pieces of clothing that would normally clash can be brought together harmoniously with the use of the right accessories. Using the latest in deep learning, they have just started decoding what they term "the genome of style."
Read the full article on the Forbes.