This article was published in Scientific American’s former blog network and reflects the views of the author, not necessarily those of Scientific American
On supporting science journalism
If you're enjoying this article, consider supporting our award-winning journalism by subscribing. By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today.
While it's possible to search the Web for images, there's still no way of searching the images themselves. Google is hoping to change this through a research project that can match digital photos of certain famous landmarks with text descriptions of those landmarks (including their namesname and where they're located) without the need for a conventional search engine.
Google created its experimental landmark recognition engine by developing a list of targeted landmarks (such as the Eiffel Tower and the Acropolis in Athens) and finding GPS-tagged digital photos of those locations. The researchers then "taught" the recognition engine to identify specific landmarks by clustering different images of the same landmark (taken in different lighting and from different angles, for example).
Here's how it might work in practice: You're browsing the Web and come across the image of a landmark that you don't recognize. You copy the image location and then paste that URL into Google's landmark recognition engine. If the image is a match with one of the landmark images in Google's database, the recognition engine returns a results page that includes the image as well as its name, location and possibly even a description.
The system is 80 percent accurate when it is given an image and asked to describe it, says Jay Yagnik, Google's head of computer vision research. Google is presenting this research, conducted with the help of researchers at the National University of Singapore, today at the IEEE Computer Society Conference on Computer Vision and Pattern Recognition conference in Miami.
Image © Google