I was thinking about this image recognition software that were talking about today in the presentation about the problem of cataloging graffiti and how this would work in a digital library. Then I realized this website is doing something similar and I love it. Its called FFFound! http://ffffound.com/ and it takes images from all over the web and posts them. There is very little text or metadata associated with them but when you click on an images it offers other images you might like, I think they must use this same software. Pretty cool.