![]()
Researchers at Carnegie Mellon University have plugged in to Flickr to create a database of over 6 million images tagged with geographic and text information. Using their software, they could take a random picture and identify where it was taken, give or take 200km, at a rate 18 times better than chance.
To get a computer to approximate this capability, the authors needed to provide it with a large database of outdoor scenes that have geographic information associated with them. That's where Flickr came in. At the time of the research, it…










