bowling.jpg

Not quite there yet but O so much closer. Google has very quietly introduced a face recognition algorithm to their image search. No,no,no, it doesn’t recognize who is in a photograph yet. However what it does, it recognizes if there is a face in the photograph. Baby steps.

Here is how it works. In Google Image, do a search on any subject you would like. Type in “blue” for example. You get as a result a series of blue skies and in your url something like this :

http://images.google.com/images?hl=en&q=blue&btnG=Search+Images&gbv=2

all you have to do now is add “&imgtype=face” to the end of the URL and your result will only show faces, human faces.

like this:

http://images.google.com/images?hl=en&q=blue&btnG=Search+Images&gbv=2&imgtype=face

Obviously, since there is no options on the interface to generate this filter, it is not ready for launch yet. And all it currently does is recognize the shape of a face, not who is in the picture. That is still in the works but not very far from being ready.

A life saver for editorial photo agencies, a complete face recognition system with identification would reduce the key wording time to almost nothing and accelerate time to desktop.
Since Google sets the standard in search, you could expect image buyers to soon come to your website expecting the same ability.

from : Google Blogoscoped

NEWS UPDATE :

Bob Roberts who had more time to play with it adds:

I tried changing “yellow” for “blue” in your posted link and got a result
that gives an idea of just how close it is,
and how far they have yet to go:

http://images.google.com/images?hl=en&q=yellow&btnG=Search+Images&gbv=2&imgtype=face

or try “queen” to see it really be useful

better still substitute “martin+luther+king” for “blue” and you can really
sense the potential.

Share Button

Comments are closed.

Post Navigation