A Swedish company, Polar Rose, is releasing a plug-in for Firefox and Internet Explorer that will allow Internet users to search for images of faces  based on visual cues, rather than the surrounding text.

Polar Rose facial recognition in action

“We are very excited by the enhanced information and linking offered by visionary search, and are looking forward to creating a whole new experience by making photos searchable in a new way”, said Nikolaj Nyholm, CEO of Polar Rose.

The company’s unique technology creates a 3D model from a single 2D image of a face. This approach radically improves photo matching by compensating for variations in lighting, facial emotions and pose.

The idea is brilliant, but Polar Rose must overcome some pretty large hurdles before this writer will be playing along.

This technology will create “facial fingerprints” on 2D images of people, and in turn, search for the same, or similar looking images on the web.

While the idea of viewing someone that looks similar to you seems entertaining, the greatest problem I can anticipate getting in the way is the fact that the engine is user trained; meaning the people who use this software will determine the accuracy of it. If this plug-in does not take off quickly, it will certainly die quickly due to the absence of users educating the database. In this case, lack of training means lack of function.

Let us not assume that the users of this plug-in will use it in a beneficial manner; you know, as well as I, that the “evils” out there will make a valid attempt at tagging popular faces with inappropriate names using methods very similar to “Google bombing.”

I’m extremely selective when installing browser add-ons. I don’t feel that there is the need for 99% of them, and I’d rather make full use of my monitor without having 320 x 200px left on my browser after the many “toolbars” have taken their share of space all while retarding my browser speed.

While the API released with this plug-in will probably provide us with more useful  3rd party apps, I for one will stick with text based tags when performing my image searches.

Related:

Entry was posted on Wednesday, December 20th, 2006 at 6:06 pm under Internet, Technology, Web2.0.

Read more entries by George Gardner.
Stumble It!