JOIN USResearchers are developing a ‘smart closet’ that has the artificial intelligence to suggest occasion-based and colour-appropriate outfits.
The Magic Closet is a computer programme, under development at the National University of Singapore and the Chinese Academy of Science.
The motion-controlled programme suggests suitable outfits for different occasions, TechNewsDaily reported.
Although the Magic Closet is a lab project, it’s almost ready for store shelves, Si Liu and Shuicheng Yan, computer scientists at the National University of Singapore, wrote to the website in a joint email.
“This kind of system is ready to use in the market,” Liu and Yan said.
“The Magic Closet can be used as a mobile personalised clothes management app. It can also be used as a plug-in system in online shops (eg, amazon.com, ebay.com) to help customers to choose suitable clothes,” they said.
The closet software makes outfit suggestions for 10 different occasions, including weddings, funerals, work and dates.
It can also match clothing to an item the user already knows he or she wants to wear. The software draws clothing suggestions from both the user’s own wardrobe and from online shops.
To figure out rules for the Magic Closet to follow, Liu, Yan and their team gathered more than 24,000 photos of outfits from online shopping sites and photo-sharing communities such as Flickr.
The scientists looked for photos that were highly rated by other people, in hopes of catching especially fashionable combinations.
They also asked people on Amazon’s Mechanical Turk, a website where people get paid to finish small tasks online, to match their photos with occasions and such keywords as “V-shape collar” or “plaid pattern”.
The team then wrote a computer programme to analyse the tagged photos, looking for rules the scientists could add to the Magic Closet’s recommendation system.
The programme could still use some work before it becomes a commercial product, Liu and Yan said.
The researchers want to develop more sophisticated matching rules. They also want to improve the Magic Closet’s ability to detect users’ bodies.
Currently, users may glam in front of the camera in different poses and the clothes they see on-screen will follow them.
That only works if they face forward, however; if they try to get a side view, the programme doesn’t know where to overlay clothes on-screen. Liu and Yan said their team will work on fixing that.