Facial recognition app for dogs helps reunite lost pets with worried owners

Eight facial markers are used to identify dogs uploaded to the app's database

Huskies pant during a training session at Feshiebridge, in Aviemore, Scotland January 23, 2013. (REUTERS/David Moir)

A new facial recognition app is being trialled by American animal shelters as a way of reuniting lost dogs and their owners.

San Diego’s Country Animal Services has already used the website and app combo FindingRover.com to reunite one local family with their Shiba inu Roxy.

The service, launched last year by founder John Polimeno, keeps a database of dogs’ faces uploaded from around the country. Eight different facial metrics are used to identify the animals, including the size of their eyes and the distance from their eyes to their nose.

Polimeno says the app is slowly being taken up by a nation-wide network that includes shelters, vets and dog organizations, all of whom will help boost the reach of the database.

It’s already worked for the Cox family of San Diego county, whose dog Roxy ran away from home when spooked by a thunderstorm. The family’s 10-year-old daughter thought to sign up for the app, and after uploading a picture of Roxy to the website soon found a match at their local shelter.

The FindingRover app lets you drag facial markers onto a picture of a dog to identify it.

"Within four hours of her arrival to the shelter, we were there to pick her up," Joanna Cox told the Associated Press.

Daniel deSousa, the deputy director at San Diego Animal Services, says that the service works whether dogs are picked up by strangers or by the county; whoever finds the animal just snaps a picture, uploads it to the database, and if there’s a match the owners get notified.

It's not the only new technology to use facial recognition either - earlier last month a new automatic cat feeder launched on Kickstarter which uses built-in cameras to check that individual felines are getting the food they need (and no more).

The technology behind FindingRover was built by two researchers from the University of Utah, Steven Callahan and John Schreiner, who worked out the eight facial markers needed to identify dogs (far fewer than the 128 markers on humans).

Callahan says that despite this, dogs’ faces are far less “uniform” then humans’, with the location of their eyes and snouts differing massively thanks to the shape of the muzzle, the breed and other factors.

This means that in a database of 100 dogs a top three match would come up with the correct 98 per cent of the time – odds that any worried family would gladly welcome to be reunited with a lost pooch.