Google is pretty smart. Sometimes Google even seems too smart. The folks at Mountain View are constantly trying to make Google as smart as it possibly can be. Google Images is a big part of the search giant. You can find an image for just about anything, and even use images to search. The next big feature of Google Images could be the ability to guess where a photo was taken.
Google has built a neural network called PlaNet. The network has been loaded with over 90 million photos that have been tagged with locations. All of these photos give PlaNet a database of information to reference. When you search for the location of a photo, PlaNet uses the database to compare photos at the pixel level. If a match is found, you’ll get a guess of where the photo is from.
Google still has a lot of work to do before PlaNet will be reliable. The first batch of tests yielded just a 28.4% rate for country and 48% success rate for the continent. Obviously, it’s a lot easier to get a continent guess correct, even if you’re not a computer. The numbers aren’t very impressive…yet. The technology will get better over time. Is this something you would use in Google Search?
[via TNW]