Polar Rose: Distributed Image Search

Polar Rose is announcing Tuesday that it plans to release Firefox and Internet Explorer plug-ins that make use of its facial recognition technology early next year. The company hopes to build an index of shared photographs online, matching faces with labels provided by its users.

Malmö, Sweden-based Polar Rose’s technology is based on creating a three-dimensional model based on a two-dimensional picture. This is meant to eliminate some of the problems encountered when comparing pictures of the same person with different lighting or with varied poses.

We haven’t been able to test Polar Rose for accuracy, and we wish they would just put out a beta already, but we do like the model they’ve chosen. Users download plug-ins and use them across any site with photos. Once a person is in the system, users will be able to do things like subscribe to an RSS feed of photos of a certain face that updates whenever new ones (even if they’re not explicitly labeled) are added.

By existing in the context of other sites, “the core functionality is photo-sharing, not photo-labeling,” says Polar Rose CEO Nikolaj Nyholm. We see a lot of plug-in companies that don’t have much in the way of technology, and we see a lot of interesting technology that we never use because it requires learning a new piece of software. Polar Rose (if it works!) will be the best of both worlds.

The two-year-old company has raised $5.1 million from Nordic Venture Partners, as we previously reported. It will release free APIs alongside its launch.

People have been trying the gamut of business models for facial recognition. Riya first went with photo organization, then said it would search the whole web, and is now focusing on products with Like.com. Nyholm explicitly said his company had gained from Riya’s efforts to democratize computer vision, as well as its missteps in over-emphasizing the technology.

That’s not the end of it. VentureBeat recently covered CogniSign, a company we like but have never written about, which has built similarity-oriented (versus direct matches, which can be quite hard) visual search that it hopes to license. TechCrunch did a piece today on Ookles, which hopes to cut it as a destination photo organization site.

And earlier this year, Google bought Neven Vision, which monetized image recognition through applications for security, authentication, and mobile marketing.

Nyholm said his company plans to make money with text advertising and a premium mobile (MMS) service. Funny to run into an internet sector where that’s not run-of-mill!