Polar Rose's facial recognition feature correctly identified half of the photos in this batch of Flickr images.
(Credit: Screenshot by Stephen Shankland/CNET)Without any big fuss, a face-recognition feature has been added to Flickr.
The new feature was launched recently by Swedish start-up Polar Rose. It lets users import all their photos from a Flickr account to an account on Polar Rose, where the images are then automatically assembled into groups dedicated to various individuals.
As with similar features in Google Picasa and Apple iPhoto, names eventually show up next to faces in the photos once the user has identified the faces. The labels then get sent back to the Flickr account. Polar Rose, founded by Swedish mathematician Jan Erik Solem in 2004, intends to license its technology to numerous Web sites.
"No other company wants to offer its face-recognition technology to all other sites," said Solem, now CTO of the company.
Polar Rose is also ready to import photos from Facebook accounts, but there's a snag.
"Facebook has a rule that downloaded data cannot be stored more than 24 hours," Solem said.
And since thumbnails are stored in the user's Polar Rose account, the start-up won't immediately be applying the feature to Facebook photos.
Already in place, though, is the authentication function Facebook Connect, which lets users log in with their credentials from Facebook. A friends list can also be imported from Facebook and can be used when identifying faces in photos from Flickr.
Solem won't say whether there's a commercial deal afoot with Flickr. "Of course we talk with them," he said.
Flickr, for its part, has kept a low profile on the subject.
"Flickr has the second most popular API on the Web," said a representative for Yahoo, which owns Flickr. "Polar Rose is one of the many third-party developers using the Flickr API to innovate and present public Flickr data in new and unique ways."
... Read more
I finally got access to Polar Rose (news story), an image search site I've been looking forward to trying. This early version of the product isn't terribly compelling, though. Today, Polar Rose can pick out faces in photos. It doesn't tell you who's who, though. The idea is that Web users who have the Polar Rose browser plugin do that.
Polar Rose plugin users can ID people in pictures by clicking on the rose icon. The icon automatically pops up below the head of of every person in a picture.
On the other side of the product is a search engine for pictures of people. If you want to find a picture of someone, and that person has been tagged by users, Polar Rose will show you all the pictures that match.
Future versions of the product will actually do facial recognition, I'm told, although the company isn't disclosing when that function will go into beta. This is the holy grail of visual search, of course. Last year, Riya demoed a tool that was supposed to do what Polar Rose is still building, but Riya eventually gave up and moved into image search for commerce (it finds shoes that look alike, for example).
I hope to see Polar Rose's technology, once it is fully developed, if it's fully developed, inside a people search engine like Spock or Wink. That's where it's needed.
See also: Google image search recognizes faces and Recaptcha: The smartest way to deal with something annoying.
You can search for people that other users have tagged.
(Credit: CNET Networks)
- prev
- 1
- next
