imgSeek (http://www.imgseek.net/) has been doing it for years and its open source! There has been some work on a web version too. |
Interesting. I checked the Retrievr about page, and imgseek is indeed credited:
"retrievr is based on research conducted by Chuck Jacobs, Adam Finkelstein and David Salesin at the University of Washington: Fast Multiresolution Image Querying (1995).1
I first came across the algorithm when someone (I think Edd Dumbill) linked to imgSeek a couple of years back; imgSeek is a standalone image management application that incorporates that algorithm as well.
retrievr is a new implementation in pure Python (and a host of great libraries: Fredrik Lundh's PIL and aggdraw as well as numarray). The frontend has been created using Helma and Macromedia Flash. " Christian Langreiter, 2006-01-02 http://labs.systemone.at/retrievr/about |
If someone wants a gigantor chunk of photos (and has an > 100gb hdd lying around for uncompressed) here's 75GB of pics from Wikipedia in *cough* one convenient file.
http://download.wikimedia.org/images/wikipedia/en/ |
IBM did this long before, look here http://www.hermitagemuseum.org/fcgi-bin/db2www/qbicSearch.mac/qbic?selLang=English
It's part of their DB2 image extender. |
It's a nice trick, however when I tried it, the results were not even close to what I expected. May be cause of my bad painting :(. |
Brian, I better start the download now so I can do something with it in 2007 :) Seriously, this is a cool pointer. The images might come in handy for specific projects. |
and don't forget about 3d search: http://shape.cs.princeton.edu/search.html |
Philipp, it's very nice they are producing tarballs again. We were downloading them 1by1 for the longest =) |