Young women going shopping

Machine learning system dishes out fashion advice

Image credit: Dreamstime

Researchers at the University of Texas at Austin, working with researchers from Cornell Tech, Georgia Tech, and Facebook AI Research, have developed an AI system intended to assess photographs of outfits and offer fashion tips.

The tool, which has been named Fashion++, uses visual-recognition algorithms to analyse the key features of garments captured in a photograph, including colour, pattern, texture and shape. It then ranks small tweaks according to how much impact they could have on the overall appearance of the outfit, before presenting several options.

Suggested changes could include switching to a longer jacket or picking a sleeveless top.

“We thought of it like a friend giving you feedback,” said Professor Kristen Grauman, a computer science expert specialising in computer vision. “It’s also motivated by a practical idea: that we can work with a given outfit to make small changes so it’s just a bit better.”

Machine learning algorithms are trained using large datasets: in this case, photographs of both well-styled and poorly styled outfits. The researchers collected 10,000 images of outfits shared publicly online and mixed images of outfits to create mismatched composite images to form the ‘unfashionable’ dataset. This information was used to train the algorithm to recognise what makes a good outfit.

Graduate student Kimberley Hsiao acknowledged that fashion – by its very nature – keeps changing, but that Fashion++ could continue to learn by being trained with new images.

Fashion++ is an imperfect tool; among other issues, it does not tend to recognise vintage looks as being stylish on account of the training images used being overwhelmingly focused on recent styles of clothing popular in the social media age. The tool also has a bias towards the type of clothing popular in North America.

Next, Grauman and Hsiao hope to train Fashion++ to recognise that different body shapes are flattered by different types of clothing. “We are examining the interaction between how a person’s body is shaped and how the clothing would suit them. We’re excited to broaden the applicability to people of all body sizes and shapes by doing this research,” Grauman said.

Emerging technologies such as AI, augmented reality (AR) and virtual reality (VR) are likely to play a large role in the future of fashion retail. The 2016 Future of Shopping report – which suggested how the ‘fourth industrial revolution’ could affect retail – predicted that VR shopping could threaten bricks-and-mortar retail by 2050, that drone delivery could become widespread, and that AI assistants could suggest purchases based on your preferences.

Sign up to the E&T News e-mail to get great stories like this delivered to your inbox every day.

Recent articles