There’s a potentially powerful new tool to help reunite pet owners with their lost cats and dogs — facial recognition technology.
The Baltimore Animal Rescue & Care Shelter announced Thursday that it is the first organization locally to use sophisticated computer algorithms that can let pet owners know almost instantly whether their missing Fido or Fluffy is at the organization’s shelter or in its network of foster homes.
John Polimeno, the California retired construction executive who founded the free service Finding Rover, said that he spent a year working with specialists at the University of Utah to develop facial recognition software to analyze the faces of dogs and cats, which are more difficult for the technology to identify than are the faces of their human counterparts. Finding Rover’s software program measures 138 features of an animal’s face and compares it with a database of photos of pets that have been picked up or reported missing.
When a front facial photo of a pet was analyzed during the testing process, Polimeno said, the software picked the correct animal out of a database of 25,000 other pets 98 percent of the time.
“If the photo is to one side or is taken from above or below, the accuracy goes down to maybe 90 percent,” he said. “We’ve helped reunite owners with more than 15,000 missing pets since 2013.”
BARCS spokeswoman Bailey Deacon said she thinks the new technology can significantly increase the number of happy homecomings between lost dogs and cats and their frantic human companions.
“Reuniting lost pets with their owners is a really tough problem that we’ve been trying to crack for a long time,” she said. “In Maryland, pets brought into a shelter only have to be held for three days before they can be placed for adoption. Sometimes, it takes owners more than three days to get to the shelter.”
Finding Rover works like this:
When a dog or cat goes AWOL, its owners can upload a photo of the missing animal onto a smartphone app. They’ll be asked to fill out some details, including the pet’s unique physical characteristics and the owner’s contact information. Owners will then receive photos of all rescued animals fitting that description, starting with the closest match and ranked in descending order — along with the contact information for the person or organization where each animal is residing.
Polimeno said the facial recognition technology even works for purebred dogs and cats which can look so much alike that even their owners have trouble telling them apart.
“Our software can recognize your dog or cat even better than you can,” Polimeno said.
He came up with the idea for the program several years ago when he noticed a missing pet poster while he was sitting in a California coffee shop with his wife. That sparked memories of the traumatic week years before when Harley, the family’s beloved black Labrador retriever, ran off and was missing for several days. Polimeno remembers driving around town with his children sobbing in the back seat of the car, knocking on doors and putting up posters. Luckily, he found the dog a few days later.
Wouldn’t it be great, Polimeno wondered out loud that day in the coffee shop, if there was facial recognition technology that worked for animals?
“My wife looked at me and said, ‘That’s how you’re going to spend the rest of your life,’ ” Polimeno said.
Facial recognition technology for humans has been around in rudimentary form since at least the 1960s. But dogs in particular posed more vexing technical challenges because they have muzzles of varying lengths that thrust out into three dimensions. Human faces, in contrast, are relatively flat and can easily be perceived in two-dimensional photos.
“It sounds counter-intuitive, but the more similar faces are, the easier it is to do facial recognition,” Polimeno said. “It’s much easier to identify humans than it is to identify dogs.”
He acknowledged that though Finding Rover’s record is impressive, the service has limits:
For instance, the database is set up for now only to analyze the photos of dogs and cats. Owners of other animals, such as horses, rabbits or ferrets, are out of luck.
And while facial recognition technology excels at identifying animals that have been rescued, owners of cats or dogs still on the loose might be misled into thinking that their pet is at a particular shelter only to have their hopes dashed. To human eyes, a photo of one Irish Setter can look much like another.
In addition, the app is only as effective as the number of pets in its database. Though Polimeno said his service has partnered with nearly 600 shelters in the U.S., Canada and Australia and that each search covers a 200-mile radius, BARCS is Finding Rover’s first Baltimore-area partner. That limits the pool of potential matches at the moment to the 300 dogs and cats being cared for by the organization.
Deacon said she hopes the pool will grow exponentially as more area shelters and individual pet owners register their animals with the no-fee service. (Expenses are paid by the nonprofit Petco Foundation, Polimeno said.)
“Is Finding Rover perfect?” Polimeno asked. “Of course not. But does it work? Absolutely. We've had amazing success stories, but it will only work in your neighborhood if the whole community registers their pets.”
Both Deacon and Polimeno emphasized that they’re urging owners to continue adopting such traditional pet-finding techniques as microchips inserted just beneath an animal’s skin and collars with identification tags or that utilize GPS location technology.
“This is an extra layer of protection,” Deacon said, “another tool for helping people to find their pets. If it helps reunite just one owner and one pet, it will be worth it.”