Do you imagine walking down the streets, taking pics of people around you and then being able to reproduce that outfit you fell in love with? It may sound c
razy, but it is just a matter of math, and it seems we are getting pretty close to achieving some actual results.
As Quartz has just revealed, the south korean SK Planet’s Lab
will launch on the market a search engine focused on the fashion world by the end of 2016. The Artificial Intelligence that they have designed is able to understand the concept of style and to match clothes according to the latest trends through scanning the huge amount of pictures that fills the net everyday.
The project has been conducted by both researchers of SK Planet and fashion designers. Together it took them over a year to develop an algorithm that can recognize thousands of different styles taken from photos on 11Street, the main online store in South Korea. When applied to e-commerce, the aim of the machine is to scan a picture, detect a style and then come up with a matching range of garments available for sale.
The peculiarity of this system lays in the fact that, while a common machine vision works only with high-resolutions, bright pictures, the new AI is designed to recognize also blurry or dark images, like most of the one taken with smartphones. This means that people don’t need to rely only on magazines shots, but can also use everyday photos to find the clothes they are looking for. The person in front of you has a mind-blowing outfit? Just take a snap of him (without being noticed, no one wants its style to be copied…) and the algorithm will do everything else.
Moreover, apart from meeting the needs and wishes of costumers providing customised experiences, the Artificial Intelligence may be also used by online retailers in order to recognise, and therefore prevent, credit card fraud in real-time.
The same idea was already presented in 2015 by a group of researchers from National Taiwan University in their study “Who are the Devils Wearing Prada in New York City?”
However they perceived their machine vision not as a business tool, but simply as a mean of showing how the clothing styles are actually influenced by the fashion trend in the New York Fashion Shows. Indeed they fir
t built a huge dataset from the NY Fashion Shows and the NY street-chic outfits, then they used the AI in order to classify specific clothing attributes (collar, panties…) and figure out which were the most popular trend and what drove them.
Now that researchers are looking to this technology from it business perspective, it seems that Math will play a major role in the Fashion industry next year: will it have the numbers to compete in this world?
by Bianca Cattadori
