Henry Kang traces the roots of his company to a rather typical moment he shared with his wife, Shawna. She was getting dressed one morning, and she asked him that all-too-familiar question: "What can I wear with this?"
Kang immediately remembered a silly scene from the movie Clueless, where the film's protagonist, Cher, faces the same question. Rather than ask someone else, she turns to her computer, using it mix and match the different pieces in her wardrobe. "I thought: 'Totally. A computer can help someone put together an outfit,'" says Kang, who carries a PhD in robotics and computer science from Carnegie Mellon University.
But this wasn't the '90s. He didn't turn to a desktop machine like the one in Cher's bedroom. He turned to the iPhone, creating a mobile app that could do the job. It's called StyleIt. You upload a photo of a shirt or jacket or a pair of pants to the app, and using computer vision and machine learning, it tells you what to wear with it.
Cher would love it---particularly because, as of this month, it also lets you instantly purchase items that StyleIt recommends, taking advantage of Apple's new mobile payment system, Apple Pay.
The app is part of a much larger movement towards mobile shopping. According to one report study, 70 percent of consumers have bought something using their smartphone in the last six months, up from 59 percent in 2013. And like many other tools, StyleIt aims to make this mobile shopping more, well, personal.
Over time, the app "learns" your preferences, much like TheTake, a recently-launched app that pinpoints products in movie scenes and lets you instantly buy them.
According to Kang, StyleIt has already curated more than 1.5 million outfits and has indexed more than 1 million items from 450 stores, including Forever21, J Crew, Tory Burch, H&M, and Urban Outfitters. It refreshes product information from this database of stores every 20 minutes.
To match outfits, Kang says, StyleIt then pulls information from fashion bloggers and sites like Polyvore. As he describes it, the app can recognize colors, textures and patterns, and based on what the users has liked in the past, it uses predictive modeling to its personalize suggestions.
Kang and his team of engineers---over half of whom are Carnegie Mellon-trained computer scientists---are confident their machine learning system can reliably match clothes to taste. After all, Kang says, it works for his wife. "She loves it. We need to take another look at our credit card bill."
