Amazon CEO Jeff Bezos is not someone you’d expect to see in the front row at New York Fashion Week, but his company could eventually be behind some of the designs coming down the runway.
Researchers at an Amazon lab in San Francisco are working on an algorithm that can learn about style from images, and recreate similarly styled garments from scratch. Meanwhile, Amazon researchers in Israel are developing a a machine-learning technique that can decide if an outfit is stylish just by analysing a few labels attached to the image. In theory, it could then recommend a better outfit, or help you choose the most stylish option between two.
The research groups from Amazon and other organisations presented their work at the KDD 2017 conference in Halifax, Nova Scotia last week, as MIT Technology Review reported earlier on Thursday. The conference focused on how computer vision, machine learning and other advanced technologies could be used in the fashion industry.
In addition to the researchers from Amazon, a group of researchers from Europe discussed ways the massive amount of fashion data available on social media sites like Instagram and Pinterest could be used to detect and predict trends and recreate the online-shopping experience.
All of this sounds pretty ground breaking, but it may be a while before these particularly techniques influence what’s in your closet. The research paper presented by Amazon’s San Francisco team, for example, only showed blurry deconstructed outfits.
But Amazon is already edging into the fashion world. In February 2016, the e-commerce giant launched seven of its own private label brands on its site. It also unveiled the Echo Look, a camera that has the company’s Alexa intelligent assistant built in. The Look can snap pictures of you and then use its artificial intelligence to help you find matching clothes for the ones you’re wearing, such as jacket to go with your shirt or shoes that would go with your pants.
An Amazon representative declined to comment on the company’s fashion-related research projects.