Shutterstock Inc. is turning to artificial intelligence (AI) to help its customers find that perfect picture.
On Thursday the New York City-based stock photo giant announced the beta release of what it calls a “Composition Aware Search” feature, which uses machine learning technology to help users search for photos not only by subject, but position as well.
And for those wondering just how much AI is capable of learning on its own, Shutterstock founder and CEO Jon Oringer said the company’s search engine taught itself to recognize where subjects were placed.
“What’s remarkable about this breakthrough is that we only trained our model to learn what things are, but our deep network learned how to represent where things are,” Oringer said in an Oct. 12 statement. “For marketers, searching for an image with copy space using this tool will save a significant amount of time.”
Since we here at ITBusiness.ca are stock photo customers ourselves, we thought it would be worthwhile to test the search engine and share the results. Our impression: while the engine can definitely help users searching for photos to illustrate the right topics, it works best when said topics are popular stock photo subjects.
Yesterday, for example, we wrote a story about personalized shopping which led to searches for shopping and mobile devices – a common theme in stock photos:
The breakthrough comes when separating search terms – for example, “mobile device” and “credit card” (separated by using commas or pressing Enter) – which produces two circles that can be moved around to prioritize photos that, say, place the credit card on the right side of the screen and the mobile device on the left:
Or vice versa:
If we were to attempt adding, say, a person to the mix, it becomes less helpful:
(Though hey, two out of four ain’t bad. Also, these are four of dozens, though only half showed a body and only four included a face.)
Still in beta, with its patent pending, Shutterstock’s new tool uses a combination of machine vision, natural language processing, and information retrieval to find matches based on spatially aware search criteria. The company’s own example was of a user searching for wine and cheese, as seen in the header image above.
Users interested in trying the Composition Aware Search feature for themselves can do so here, or watch the video below.
