What Is Visual Search?

29th November, 2016 by

One of the most prevalent elements of the internet era is “tagging”. This is how we sort content, make searching much easier, and streamline the process of finding what we’re looking for in the vast pool of content that out there on the world wide web.

These days, most content creators know that in order to create compelling and shareable blog posts and articles—or to create an appealing listing for a product or service you are selling in an online marketplace—images must be added in order to heighten impact. However, with so much content on the internet these days, wading through the thousands, if not millions, of images associated with a common tag can be a daunting task. Indeed, the time investment required for finding a suitable image in a sea of similarly tagged files can be irritating and draining. In addition, there is the issue that not all tags are created equal, so the quality of images can vary tremendously under a single tag.

There is a key solution for that dilemma which is making its way into the internet mainstream: visual search. This is a form of artificial intelligence which allows a machine to identify the contents of an image, and match both visual and text-based elements without the need for a human to manually tag it. The more photos that are sorted, the more an algorithm can learn what objects are in the photos.

The early iterations of visual search have been with programs like Google reverse image search and TinEye. These both work by “extracting specific patterns [in an image] and how they contrast with their surroundings”, and are helpful when you’re trying to find a higher res version of a photo online or trying to see other locations where a given image exists online.

However, we’re starting to see more and more major internet companies wake up to the power of visual search as a means to wade through and actually categorize vast amounts of images online. eBay recently acquired visual search company Corrigan to improve the user experience on its site. In a statement about the rationale behind the acquisition, eBay said: ““With more than one billion live listings on eBay’s platform, Corrigon’s expertise and technology will help match the best images to their products so that shoppers can be confident that what they buy is exactly what they see.”

Another company that is focusing on the capabilities of visual search is a New York based startup called Clarifai. Clarifai’s goal is to take visual search powered by an algorithm into the mainstream by empowering more people to build it into their platforms and apps. As TechCrunch reported, “While Google, Pinterest and other companies build visual search technology, Clarifai is looking to do the same but focus on giving third-party applications and developers access to that kind of technology. Zeiler says Clarifai only needs a few images’ worth of data to start building out a model for determining what kinds of objects are in an image. Developers can teach algorithms with their own kinds of tagging to build new classes of “objects” within those images and videos.”

Visual search is an exciting area of development in the AI space, and the more developers can build it into their products, the more than consumers will grow accustomed to using it. Even more exciting is that the “ability to build an understanding of data structures” has applications far beyond mere image search and categorization. Thanks to this kind of machine learning, we can see a future where the internet is much easier to move through thanks to better—and less labor-intensive—means of content categorization.

(Visited 31 times, 1 visits today)