It seems there's not as much ambiguity in the visual world as you might guess.
Estimating geographic information from images is a difficult, but very much a doable, computer vision problem.
Carnegie Mellon graduate student James Hays speaking in a CM press release about the visual pattern recognition system that he developed with Assistant Professor Alexei A. Efros which uses geotagged Flickr photos to place any geographically specific image on the map.
Hays and Efros' algorithm can accurately geolocate the images within 200 kilometers for 16 percent of more than 200 photos in their test set -- up to 30 times better than chance.
The system "analyzes the composition of the photo, notes how textures and colors are distributed and records the number and orientation of lines in the photo. It then searches Flickr for photos that are similar in appearance."
It'll be interesting to see whether the same algorithm can be applied to video.
Read the press release »
2 comments:
... did you see Blaise Aguera y Arcas' presentation on TED last year?
discovered when I went to tag it that I had already tagged it sometime back -- thanks for the swift kick to finally *watch* it -- I'll take a look. :)
Post a Comment