Reinventing The Lens

Face Recognition for Image Searching

by Ron Leach

With the Internet becoming more and more visually oriented all the time, a Swedish company has devised a unique approach for searching the millions of images that are uploaded daily. The new Polar Rose technology employs a combination of user input and facial recognition algorithms to enable you to search photos by content and identify people pictured in the images.

Still in beta form, Polar Rose uses facial recognition technology developed at the universities of Lund and Malmo in southern Sweden. Their approach is unique in that it is able to extract 3D information from 2D images--thereby surpassing the accuracy of existing approaches to face recognition. The Polar Rose method also relies upon the "collective intelligence of our users" that will help "train" the engine by tagging images with names and other annotations.

Polar Rose plans to offer two versions of their search technology, the first being a plug-in for Firefox and Internet Explorer web browsers. This plug-in will place an icon on an image that users can click to obtain the name of the person in the photo as well as other relevant information. Polar Rose says they will also provide for the royalty-free use of their programming interface so that partners can integrate this functionality into existing or future sites or stand-alone applications. News sites, for example, could utilize this technology to enable users to help tag photos and link stories based upon the identities of people appearing in the photos.

All this sounds quite interesting from a developmental point of view. However, there are some important issues regarding privacy, distribution rights and other considerations that will likely have to be addressed should this technology move from the realm of innovation into common usage.