Some new technologies will make you go “gee whiz, that’s pretty neat-o”. Others will cause the stronger reaction “gosh dang jeez, that’s simply witchcraft, get the pitchforks, torches, and holy water”. TinEye, now in public beta from Idée of Toronto, gets very close to appearing as witchcraft.
TinEye is an image search engine, but unlike per se, Google Image Search, the search query is an image, not a word or phrase. You upload an image, or supply a URL to an image, and Tineye returns images that match or partially match it. Showing all the places on the web that a particular image is used is interesting in itself, but the partial matching is where it impresses most. For instance, searching with an image of Da Vinci’s Mona Lisa returns not only various copies, sizes, and color variations of the original, but also edited parodies of it, book covers using it, and photographs of the painting itself.
Idee’s image search technology is already in use by various news agencies to track use of their photos, and by Digg to prevent duplicate image submissions. Now Tineye has opened it to public use for the first time. The site cautions that their index is still “very small”, but claims 701 million images indexed; a small fraction of the web, but a good start. With a more comprehensive index, TinEye will become a very compelling tool.
Interesting/scary thought: Tineye indexing the images of Google Street View; image search of the real world. Then we might have to break out the pitchforks.