A.I. Pet Detectives Are Here
Published on Monday, 22 July 2019
Apparently a big leap forward has taken place in the animal facial recognition space. No less than three reports have sprung forth over the past week about technological startups providing facial recognition for pets, and they are all so good that we must all but declare that this unmet need has been solved.
So prepare to say goodbye to GPS tags, dog tags, microchips, tattoos and inscribed collars.
These A.I. contenders for Ace Ventura: Pet Detective's job may not be human, but they were designed to prevent one thing first and foremost: heart break.
So how do they work?
The first of the trio, Megvii, uses nose-print recognition technology which enables biometric identification of your dog. In case you are wondering, it doesn’t require direct contact like your phone’s fingerprint scanner. Instead it works more like your iPhone’s Face ID.
Megvii’s software only requires you to take a series of images or videos of your dog’s snoot from different angles. These are then uploaded to a dedicated database and analyzed by the software to determine critical identification markers.
How accurate is it? Expect 95% accuracy, say the developers. And you wouldn’t even need to boop your pet’s snoot in ink to make a nose print.
The startup is backed by Alibaba, so expect it to hit mainstream usage in the not-too-distant future in China, possibly doing away with the need for microchipping there.
Over in America, apps like Finding Rover have apparently already found traction in over 700 shelters around the world, providing them with animal facial recognition technology linked to pet owner databases.
Finding Rover’s CEO says their database has more than 700,000 pet pictures already (as of July 2019) and is growing its network by five to seven shelters a week.
Like Megvii, all you have to do is upload your pets’ photos to the app, and the software does the rest. Finding Rover claims an identification accuracy rate of 98%.
PiP is another app that provides 98% identification accuracy rate, on par with Finding Rover – and both better than Megvii.
It however, works on a paid subscription basis, so you’ll have to pay to use PiP – not that we know if Finding Rover, Megvii or PiP are ever coming to Singapore.
Who will find these most useful?
Obviously, lost and found animal shelters, and pet-owners to start with.
Imagine being able to match pictures of lost dogs and cats with new shelter arrivals or found pets, to reunite them with their owners. It’ll be as simple as snapping photos or videos of the dog/cat – and alerts can be sent instantly to the pet owners via SMS or in-app notifications.
Maybe AVA and SPCA should get with the program and develop a national database for pets that uses these AI? After all we are supposed to be a Smart Nation.
This will definitely help prevent the culling of beloved domestic pets by accident.
In this incident which occurred around 2012, a pet-owner had contacted SPCA about her missing poodle. But before she could claim her dog, it was put to sleep.
Amongst the chain of errors that led to the tragedy:
- a pet owner's calls to the animal welfare group, claiming ownership of the 16-year-old dog, never reached the team in charge of euthanising animals
- the society's shelter manager had misread the letter "C" on the dog's microchip as a "6", and was therefore unable to verify that the poodle had an owner
- the SPCA gives pet owners four days to claim their lost pets. The poodle's owner had spent the first 3 days looking for her dog, and another day retrieving her dog’s microchip number from AVA. By the time she was able to confirm her ownership of the dog, it was already too late
- the SPCA still has no plans to extend the four-day time frame.
SPCA has since modified procedures to prevent similar occurrences.
But implementing AI pet recognition technology could eliminate up to 98% of all future occurences. For all time.
And the only posters you would be putting up would look like this instead: