The written language, made of letters, words and phrases, is how we mostly do internet search. It doesn’t have be that way though. It is very possible, for example, to search a photograph by drawing its approximation. This is shown by ‘retrievr’, the graphic search engine of System One Labs. In retrievr, you can do exactly this type of non-textual searches. You can also search by uploading an image. ‘Retrievr’ searches for results in the flicr database.
‘Retrievr’ is a python implementation of an image search algorithm originally developed by Chuck Jacobs, Adam Finkelstein and David Salesin at the University of Washington. This algorithm is also implemented in imgSeek is a standalone image management application for UNIX systems (such as Linux or Mac OS/X).
It is interesting to put search engines such as retrievr in context with other non-textual search applications : Shazam, for example, the song recognition service on iPhone and Android. For what we face today is only a biginning of the development of search engines of this type. Projects like these will most probably contribute to the ongloing decline of the importance of text in favour of images and sounds, which has been on its way ever since the advent of television. Step by step, we are entering, perhaps, a post-textual era.
The TinEye search engine allows you to find images on the web based on an image-file URL.You can also upload the image to search for.
