Home furnishings retailer Wayfair Inc. has launched a new site search tool that allows shoppers to search for products with a photo.
Here’s how it works: Consumers tap the camera icon in the Wayfair search bar on its desktop or mobile site or in its app. From there a shopper can take a photo of product she wants or upload one from her photo storage. Wayfair will then match the image with its 8 million products and surface the results.
The feature aims to save shoppers time and help them find what they are looking for faster.
“We know that everyone has a unique vision for their home and that ideas and inspiration can come from anywhere. A sofa in a friend’s house, a Pinterest board, a barstool at a favorite restaurant, or a pillow featured in a design magazine, can all spark ideas for decorating one’s home,” says Wayfair co-founder Steve Conine. “While finding a look you love may be easy, finding the specific products to bring that look to life can be time-consuming and require a lot of searching.”
The visual search tool uses a machine learning algorithm so it will improve its search results the more consumers use it. The feature is part of the retailer’s propriety computer vision system, says Matt Zisow, director of product at Wayfair.
“We developed that system using a deep convolutional neural network approach that others, including Google and Pinterest, have used to develop similar applications.” Zisow says. “What sets our application apart is the rich, massive, proprietary furniture and decor data set that we used to train our models, which continue to get smarter as users engage with visual search results.”
A convolutional neural network is a machine learning technique modeled after the brain structure that often is used for visual recognition. Such a network would be exposed to images of lamps, for example, and then be able to automatically identify other, similar lamps, even if it had not been exposed to images of them.
The tool took about three months for Wayfair to develop, Zisow says.