Polar Rose Promising Face Recognition Image Search

I got a press release today from Polar Rose,
a new company promising to bring facial recognition technology to those seeking
images from across the web. In other words, want to find all the pictures of
Bill Clinton? Polar Rose says it will make it happen through a browser plug-in
that works with existing photo services, as well as through partnerships. If
this sounds familiar, I’ll revisit Riya’s recent foray and back away from facial
recognition.

The full press release is below. In short, Polar Rose says that sometime between
January 1 and March 31 of next year, they’ll release a browser plug-in. Install
this, and you can then apparently tap into image search engines such as Google
Images or Yahoo’s Flickr and bring back results that match the faces of the
people you are looking for.

TechCrunch has a nice

screenshot showing what I assume is the plug-in being used to tag faces in
Flickr for within the Polar Rose service. It doesn’t really show how the service
will reshape results at a place like Flickr.

Facial recognition is said to be done by making 3D models of faces spotted in
images. In addition to automatic matching, the system will also use user
training. This will likely be similar to what’s shown in the screenshot, where
people are asked to manually tag faces. Get enough people tagging faces, and
that can be added to improve the algorithmic guesses that are made.

Of course, this story sounds similar to Riya. That service
kicked off
in March of this year, asking people to upload photos and tag them, to build a
facial recognition database. Soon after, Riya

changed to become a general web image search service, which remains
operating at Riya.com. But most efforts
shifted again
in November to focus on image recognition tied to fashion products via the
Like.com site.

Before Riya launched, Google was
rumored to be
interested in acquiring it. That

didn’t happen, and Google
later acquired
another image recognition company, Neven Vision. Google has yet to roll anything
out approaching facial or image recognition, however, simply

promising general features to come.

The most substantial image matching service Google has released recently was
the Google Images Labeler
game, which I’ve covered more
here. Aside
from that, Google

added search features to
Picasa
Web Albums this week, including I believe the new ability to tag photos.
Picasa Web Albums have only been

offered to the general public for three months (and a more limited group
since
June), and Google is clearly rapidly shaping the service up to challenge Flickr.
But image recognition, despite the Neven Vision acquisition, has yet to roll in.

That brings us back to Polar Rose. Despite literally years of excitement and
anticipation over facial recognition, so far no one has had a consumer hit with
it. Riya’s attempt clearly didn’t catch on. Perhaps Polar Rose will have more
luck. Aside from browser plug-ins, the company also hopes to convince partners
to embed its technology into their services. Few search technology companies
have had much success with the "add us in" approach, so that’s another big watch
and see. Even larger will be the actual beta launch. I’m always wary of
announcements before anyone can actually play with the technology. Let’s get the
beta out there released to anyone, then we’ll all be better able to tell how
useful Polar Rose really is.

Here’s the press release:

Polar Rose Launches Visionary Search for Online Photos

New Technology Enables Advanced Face Recognition for Every Photo on the
Web; Automatically Enhances Photo Search and Web Albums to Create a New Way of
Discovering Information

December 20, 2006 – Malmö, Sweden – Polar Rose, a company uniquely
combining automatic face recognition with 3D modeling for visual search
applications, today announced its product and an open beta trial, beginning in
Q1 of 2007 at https://www.polarrose.com.

Using Polar Rose’s technology, it will now be possible to automatically
search for and recognize faces on a consistent basis similar to the way people
do – by using the visual cues of the photo as opposed to nearby text analysis.
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.

Available as a free Web browser plug-in and through royalty-free APIs
(Application Programming Interface) for partner integration, Polar Rose will
seamlessly enhance the experience of browsing and searching for photos of
people. Polar Rose, which is also the name of a flower-shaped mathematical
curve used as a logo mark by the company, will augment any site with true
people search and enable linked data discovery in connection with a person in
any photo.

A search for “Peter Jackson