Photo by Joanna Cox
Roxy, a 10-year-old shiba inu, was lost for five days last summer when she bolted during a thunderstorm. Owner Steve Cox credits a facial-recognition app with bringing her home.
To the imperfect world of lost-pet identification comes a new technological approach: facial recognition.
Free apps offered by two competing companies in North America aim to identify and recover lost dogs and cats. Called Finding Rover and PiP My Pet, the computer- and smartphone-dependent apps aren’t meant to replace microchips, tattoos or collar ID tags, but supplement them.
“Do all of the above,” advised Colleen Hodges, a spokeswoman for Harris County Veterinary Public Health in Houston.
Traditional pet-identification systems are notably fallible. Tags on collars can fall off or become illegible. Pets can slip their collars. Microchips have myriad problems: Not all scanners can read all chips; and by themselves, chips tell only where an animal’s identity may be found. Chips must be registered, and there are multiple registries on which the subject animal’s information may be located — assuming it’s registered at all. Tattoos, like microchips, must be registered and the owner’s contact information kept up-to-date to be effective.
Applying automated facial-recognition technology to finding lost pets is an attention-getting undertaking. Finding Rover, the first of the two companies to enter the market, says it already has 100,000 pets registered in the United States, Canada and Australia, and more than 200 shelters signed on or in the queue to become partners. Companies whose focus is pets, including VCA Inc., one of the largest owners of veterinary clinics in North America, are taking an interest, as well.
But in the same way that business interests, patent protections and politics have left themicrochip-ID system in fragmented disarray, competition between facial-recognition companies using proprietary algorithms and keeping their own databases of registered pets could limit the utility of the approach.
The developers acknowledge the risks of maintaining separate systems but at this early stage, as they establish and refine their products, neither is ready to collaborate.
“I’m in favor (of), you pick a horse and you ride it in a race,” said John Polimeno, founder ofFinding Rover.
PiP My Pet co-founder and CEO Philip Rooyakkers has a similar view. “There’s a reason you patentthings…” he said. “Do you make (the technology) open source at some date? Yes, you certainly could. But in the beginning, as you’re trying to create this, you protect it, right? You don’t just give it out to everybody and anybody.”
Competitive issues aside, proponents of facial-recognition technology don’t claim it to be the perfect solution to the difficult business of reuniting lost pets with owners. Said Rooyakkers, “I see this as another weapon in the arsenal.”
Broad participation needed for success
Judging from the experience of shelters, more tools certainly are needed. In Harris County, Texas, most of the 25,000 animals a year that go through the shelter system have no identification, according to Hodges. “None,” she emphasized. Lack of ID obviously makes it hard to reunite pets with their owners, especially with limited time.
Many shelters hold found animals for only 72 hours before they’re euthanized or become available for adoption, according to Rooyakkers. “Speed is critical,” Hodges affirmed. “People sometimes don’t realize that.”
Her community’s public shelter in February became the first in the United States to integrate its data with PiP My Pet, which is based in Vancouver, Canada. (PiP stands for Positive Identification of Pets.) Photographs of all animals entering the shelter are sent to PiP My Pet to allow for automatic matching with lost pets reported via the app.
Other shelters have integrated their systems with California-based Finding Rover. At the San Diego County Department of Animal Services in California, one of 77 shelters partnering with Finding Rover, Deputy Director Dan DeSousa is enthusiastic about the potential for the app to prevent lost dogs from ever entering the shelter.
“If … my dog is missing, I can report it (to Finding Rover) as lost, and it puts out an email alert to everybody in the area that has the Finding Rover app,” he said. DeSousa added, “We need the public’s buy-in and the public’s participation.”
VCA Inc., which owns more than 600 veterinary clinics and other pet-related businesses, recently signed on as a Finding Rover partner. Exactly what the partnership will entail is still being worked out. For now, Jacinthe Moreau, VCA marketing director, said only, “We’re excited about the technology and what it represents for animals.” Like DeSousa, she noted that the more people hear about and use it, the better it will work.
How does facial recognition for lost dogs work?
California pet owner Steve Cox found out after he and his family searched for days to locate their 10-year-old shiba inu, Roxy, who escaped the house without her collar during a thunderstorm last July in the El Cajon area of San Diego County.
Roxy’s family posted fliers and visited animal shelters and shelters’ websites. No luck. But at a county animal shelter, Cox came across a Finding Rover brochure. He took it home, where his 10-year-old daughter read it and followed the instructions.
She rounded up the app, registered and uploaded a recent photo of Roxy.
That was on a Sunday. The shelter emailed Cox two days later with a photo of a dog a staff member had taken and submitted to Finding Rover. It was Roxy.
“We were elated, crying and very happy when we saw her picture,” Cox said. “It had been five long exhausting days, and we started thinking we may never see her again.”
Roxy had arrived at the shelter just that day. When the Coxes came to pick her up, she’d been at the shelter only four hours.
Cox is a big supporter of Finding Rover now, recommending the app to his friends. Roxy has become something of a poster child for the company, which features her on its website.
Better than a winery
Roxy is one of more than 480 dogs reunited with their owners via Finding Rover since the app debuted in August 2012, according to company founder Polimeno.
Polimeno is a retired construction-firm owner. In a telephone interview, he said the idea for the app came to him when he noticed a lost-dog poster while sitting in a coffee shop with his wife, Kristie. The posting brought back memories of their dog, Harley, going missing years ago. Their kids were in tears until Harley turned up a few days later.
In the coffee shop, Polimeno made a remark along the lines of: “They can ID people with facial recognition. Wouldn’t it be cool if facial recognition could be used on animals?”
Kristie agreed. At the time, the couple were planning to develop a winery on their Northern California property. Kristie loved the facial-recognition-for-pets idea so much, she urged him to use the money set aside for the winery on that, instead. “There are a lot of people who are probably going to make wine better than you do,” Polimeno remembers her saying, “but you could be remembered for saving thousands, or hundreds of thousands, of animals.'”
Polimeno hired a team at the University of Utah Software Development Center to develop what is now Finding Rover’s patent-pending facial-recognition technology.
128 points on face compared
Steve Callahan, who with his colleague John Schreiner developed the Finding Rover algorithm, explained the tool this way: “The algorithm looks at a collection of dogs and determines what the most distinguishing features are of all the images. Those features could end up being stripe patterns, distance between eyes, texture, color, etc. It then uses those important features to compare the search dog with all the others to see which match is best.”
Polimeno said the algorithm considers 128 points on each subject’s face to liken to or distinguish it from others. Users are instructed to submit a recent photo of the subject animal looking straight at the camera, with the face clearly visible. But because the app considers so many aspects of a face, it is somewhat forgiving. For example, in Roxy’s case, the family used a picture that was handy, showing the dog wearing sunglasses. Although Roxy’s eyes were covered, the app had many other points to search for comparisons.
To make the search less onerous, the app initially narrows the field by comparing photos of lost animals against animals found in a 10-mile radius. When someone uploads a photo of a stray or missing dog, the app checks hourly for a possible match at local partner animal shelters and with dogs found by participating individuals.
Polimeno said the person who lost or found the pet may expand the search up to 2,000 miles.
How accurate is the match? Polimeno said repeated tests during development honed the app to be 99 percent accurate for cats and 98 percent accurate for dogs. This means, by his explanation, that given 100 dogs in a database, a photo of the target dog will appear as one of the top two of all possible matches.
In the real world, that result presumes that a photo of the lost dog has been uploaded by a finder, whether an individual or a shelter.
Roxy’s owners were lucky. She landed in a shelter that had been integrated with Finding Rover’s database.
A push to partner with shelters
The vast majority of animal shelters aren’t integrated with either pet facial-recognition company. Polimeno said he is aiming to sign as partners large municipal shelters, which entails pursuing approvals from city councils and county boards of supervisors. “It takes a long, long time,” he said. “It’s not like going to a humane society that’s private.”
Rooyakkers at PiP My Pet, which marked a “soft launch” in mid-2014, said his company is integrating with six shelters in Los Angeles County, the Houston SPCA and Silicon Valley Humane Society. “We are in discussions with animal-control agencies and rescues across the U.S. and Canada,” he said.
With either company, a shelter partnership entails connecting through software; it involves no exchange of money.
While neither company expressly prohibits partner shelters from integrating with both pet facial-recognition systems, they’re not keen on it, and partner shelters seem disinclined to ally with more than one, at least for the time being.
Said DeSousa at the San Diego County Department of Animal Services, which is partnered with Finding Rover: “PiP My Pet is very new. Right now, we’re going to stick with Rover. They’re the big dog in the pack.”
Hodges in Harris County, Texas, PiP’s first partner shelter, said that while no technical or contractual barriers prevent the shelter from using multiple systems, “It’s more a matter of our time and resources.”
She said her agency chose PiP for its versatility. “We chose to go with the app that we felt was specifically developed and designed to perform facial recognition of both dogs and cats and seemed to have the better algorithms,” Hodges said.
(Polimeno said Finding Rover can match cats, too. Technically, the app doesn’t know that it’s looking for a dog — or for that matter, any animal. For all it “knows,” the tool could be comparing oranges. Polimeno said his team decided to concentrate first on dogs, and that finding felines will become more of a focus for the company possibly starting in July.)
PiP My Pet’s basic approach is broadly similar to Finding Rover, with some differences. The technology, developed in-house, works only on iPhones and the Android mobile operating system, so users need a smartphone to sign up. PiP also asks for information such as a secondary contact in case users aren’t available when their lost pet is found.
Another distinction from Finding Rover: Rather than allow pet seekers and finders to contact one another directly, PiP employs a customer-care team to contact an owner when the app finds a match. The team makes arrangements for a reunion at a pre-arranged location, such as a veterinary clinic.
This way, “You know there’s safety for everybody involved,” explained Rooyakkers.
Since its debut several months ago, PiP My Pet has reunited pets with their families five times, Rooyakkers said.
Rooyakkers has been in the business of pets for years. In 2002, he co-founded The Urban Puppy Shop in Vancouver, a dog food and supplies store that also proffers boarding, day care, grooming and walking services.
Apps to come with advertising
With their apps free to users, PiP My Pet and Finding Rover are looking to make money eventually through advertising or related marketing.
Registering requires users to provide a certain amount of personal information. The privacy pageof the Finding Rover website states in part: “… we may share your personal information with carefully selected business partners so we may communicate with you about special offers related to Finding Rover programs or contests that may be of interest to you. These business partners will only be authorized to use your personal information for marketing purposes related to pet products, safety and health, and contests.”
Opting out of disclosure of personal information to third parties requires submitting a request in writing.
However, Polimeno said Finding Rover has no intention of sharing anything other than aggregate information about users with advertisers. “The model is similar to Facebook; it’s all in-app advertising … (that) we would put in their feed,” he said.
“You’re never going to get spammed by emails,” he promised.
Polimeno said the existing privacy language was “copied from another app” and will be changed.
Rooyakkers of PiP My Pet said he anticipates the venture to make money eventually by evolving into “a complete pet ownership lifestyle app” that lets pet owners book dog grooming, day-care/boarding, and veterinarian appointments directly via the app, “charging the business a small fee for those services.”
“We also believe that there will be opportunities for allowing pet-related products advertised or sold directly through the app itself,” Rooyakkers added. “But first and foremost will be establishing the technology as another tool for pet owners to help them get their beloved pets home should they ever lose them.”