First impression on unpacking the Q702 test unit was the solid feel and clean, minimalist styling.
Start-up tunes facial recognition system for online dating
- — 13 November, 2007 08:29
A start-up that built a facial-recognition search engine wants to give online daters the ability to search for mates who look like celebrities, but ultimately hopes its technology will be used for larger issues, such as helping law enforcement track down child molesters.
ActiveSymbols unveiled today a beta version of its Eyealike Visual Search platform, which analyzes facial features, such as eyes, nose, mouth, chin, skin tone, hair color, texture and length, and compares them to hundreds of thousands of photos on the Web.
Beta participants can upload their own photo to discover which celebrity they look like, or click on a celebrity they find attractive and search for daters with similar facial features. The public beta site crawls through about 300,000 images of males and females posted on such dating sites as Match.com and AmericanSingles.
The Eyealike site itself is meant to demonstrate ActiveSymbols' capabilities, rather than become a destination site of its own. ActiveSymbols hopes to incorporate its technology into that of major dating sites, but so far does not have any signed agreements to do so, says company President Greg Heuss.
"I have a long history and background in the online dating world going back to the late 1990s," Heuss says. "I've already begun conversations [with some large online dating companies]."
Heuss was vice president of marketing for Kiss/udate.com before selling the platform to Match.
Eyealike breaks down facial components and analyzes them separately:
So if you're a fan of Charlize Theron, you can look for potential mates with a similar nose, skin tone, chin shape or hair texture.
Heuss says this is more effective than simple facial-measurement tools. "There's a ton of open source code that can find a face and the distance between the eyes. But it's when you break down the thousands of pixels in the face that makes us different," he says.
The EyeAlike algorithm tries to figure out what each person is searching for, so if someone clicks on a number of blondes, people with that hair color will come up more frequently. Additional features planned for December let users search more effectively by specifying which facial features are most important to them.
The company's long-term plans include a product for government agencies in the second half of 2008.
Heuss hopes law-enforcement agencies will want to use this facial-recognition technology to search the Web to find the whereabouts of criminals, such as accused child molesters who may have posted their own photos on a Web site.
"There is a possibility. That's the furthest thing down the road for us," Heuss says.
ActiveSymbols was spun out of Logicalis this year to focus on the problem of image search, which typically relies on humans to create text associated with pictures. ActiveSymbols CEO and founder Jeff Reed was the CTO of Logicalis before his new venture in the online dating world.
Heuss laughs when asked if he's worried Eyealike will ruin online dating for ugly people.
"I've had some people ask that specific question," he says. "I don't know, I guess that's a question of the 80/20 rule, I suppose. For the dating sites, the top 20% of the people will get 80% of the search results."
Even if the EyeAlike facial recognition tool does get incorporated into sites like Match, it would be used in addition to current searching parameters, not replace them.
The start-up, which refers to its overall technology as an "enterprise search solution for facial recognition, image detection, and video copyright surveillance," says it plans to roll out products for video copyright protection, stock photography, art and e-retail, all in 2008.
ActiveSymbols says its video-copyright tool is coming the first quarter of next year, and a private beta can be joined by e-mailing firstname.lastname@example.org
Plans for the first half of 2008 also include stock photography, to help people automatically tag images; and e-retail, to help consumers look for products. However, both of these tasks are already performed by such sites as Riya and Like.com.