If the human brain sees a million images per day and can instantly identify them, why couldn't software do that, too? Making such a thing a reality has been the longtime goal of German-born physicist Hartmut Neven, whose facial-recognition software firm was purchased by Google in 2006, with the stated goal of bringing his vision to digital photography.
Instead of sifting through thousands of computer file names and badly named photo folders, what if we could organize our pictures by faces instead, gathering all of our pictures of Mom, Dad and the rest of the brood together?
Neven joined forces with Google's Picasa photo editing and management software team, spending two years developing a tool that could bring photo facial recognition to the masses.
Now the software is here, and Google is winning raves for its accuracy. It's not perfect, as Google engineers point out, but it's still a major step forward.
"I find it interesting to understand how the human mind works," says Neven, who now works at Google as a lead technical manager. "How it is so adept and capable at learning. And the best way to teach the theory on brain functions is to write a program that's quite similar."
Google's face-recognition system ties into Picasa Web Albums, Google's online photo-sharing service. It's one of the search giant's smaller offerings. Web Albums had just a 3.4% share of the online photo market in August, according to researcher Hitwise, ranking No. 5. Market leader Photobucket had a 34.7% share.
Google recently upgraded its Picasa desktop photo-editing application with faster uploading tools. Face recognition, however, happens on the Web. After uploading pictures to Web Albums, you simply click the "add name tag" feature. A group of 200 shots from your family vacation, for instance, will be broken into groups based on the faces — 60 of Dad, 40 of Mom, etc.
Just match names with faces
Your job is to match the name to the face by typing in their names. The idea is that, after each face has been identified and tagged you could use the information to, say, call up a group of photos of brother and Mom together, or you with a specific group of friends, and make an instant slideshow.
"What we're doing is taking measurements," says Mike Horowitz, the Google product manager who oversees Picasa. "We're looking at the special elements of the face — where the eyes and nose are, for instance, and trying to make a prediction."
Chris Chute, an analyst at researcher IDC, says photo recognition is something consumers will latch onto, because it solves a growing problem. "We're taking more pictures than ever before, because we can, and the ability to organize them is up to the user," he says. Since many people don't like to take the time, "Anything that makes it easier for them will be greatly appreciated."
Privacy a concern
Facial-recognition applications are a growing phenomenon. Authorities in Germany and England are testing computerized facial recognition in conjunction with new biometric passports. Passengers go through security and deal with computers, instead of humans.
But can computers be trusted? How much do we want them to know about us? In the Tom Cruise film Minority Report, a computer spits out personalized ads directed at the character everywhere he turns.
Google has been a frequent target of criticism on privacy issues, and tech bloggers have weighed in on the new face-tagging feature.
"I don't like it at all," says Rob Williams, who blogs for the Techgage website. "Google knows what I search for, where I live and how much time I spend on websites. Now they know what my friends look like, too. That's just too much."
Horowitz says the tagging feature is "opt-in," so only people who choose to use the feature do so. "Turn face tagging off, and all the data goes away."
For a scientist like Neven, there's no such thing as too much information. Before joining Google, Neven taught computer science at University of Southern California, then started his own company, which became Neven Vision.
"We were a start-up and had to make money, so we turned to Europe, where cellphones were more advanced, and worked with brand marketers," he says.
His goal was to use software to recognize billboards and other objects. He worked toward getting snapshots from cellphones to return information about the product.
In one European promotion for Coca-Cola, people were asked to take a quick camera phone shot of a Coke can for instant registration in a sweepstakes.
At Google, Neven would like to see his software advance to the point where it can eliminate many of the mundane uploading tasks camera hobbyists now encounter.
For instance, once the application learned the faces you regularly photograph (like your spouse and kids), you would not have to manually tag photos when you transfer them from a camera to your computer — the software would do that for you.
"We want to make it increasingly automatic and seamless," he says. It's not there yet — but it will be, promises Neven.
