Startups sniff up missing dogs with AI, ‘nose prints’

TOKYO — Tech companies in Japan and other countries have sought ways to reunite lost dogs with their owners, developing AI that can identify these precious puppies from nothing more than pictures of their noses.

Like a human fingerprint, a dog’s so-called nose print is unique and remains the same even as the animal ages. Tokyo-based startup S’more and others have tapped into that quality in apps that they say make it easier to track down dogs when they go missing, giving owners some welcome peace of mind.

That’s because the pet care industry is expanding in many parts of the world, with owners in Japan and other places spending a lot of money on their furry friends, especially as households get smaller.

“If our app is widely used, the concept of a stray dog ​​will disappear,” said S’more Chief Operating Officer Satsuki Sawashima. “We want to build our big data on dogs — something that hasn’t existed until now — and contribute to the development of the pet industry.”

Users of S’more’s app upload a few photos of their dog’s nose. If the pet goes missing, anyone who encounters the lost animal can submit a photo and the app should record a match, paving the way for a happy reunion. The AI ​​technology at the heart of the process has used “deep learning” to study images of 2,000 dogs and has achieved an identification accuracy of approximately 90%. A beta version of the app was rolled out in May.

S’more also plans to expand its app with features that will allow users to manage their pet’s vaccination records and other health information.

Outside of Japan, South Korean startup Petnow and US food company Mars have also developed noseprint recognition apps for pets.

And in China, the Alipay payment service of Ant Group, which is affiliated with the Chinese e-commerce giant Alibaba Group, introduced a nose print recognition system with a twist in July 202. The pet medical insurance service is using technology to verify pets’ identities to prevent false insurance claims, according to local media reports.

Meanwhile, Japanese pet insurance company Anicom Holdings has developed an AI-based system that uses facial images to predict whether dogs are at particular risk for certain diseases.

The company used facial images of toy poodles to build the system. It divided the photos into those of healthy dogs and those of dogs that developed eye, ear, or skin diseases, and then let the AI ​​learn them.

The company says the app has a prediction accuracy of about 70% for eye diseases and about 60% for ear and skin diseases.

“By increasing accuracy, we want to bet [the system] to users within two to three years,” said Ryo Horie, executive at Anicom Insurance, a subsidiary of Anicom Holdings.

Similarly, South Korean startup AI FOR PET has developed an AI-based app to help diagnose whether pets have diseases based on images of their eyes and other body parts.

Owners submit photos from their smartphones and are then informed if their pets have any disease-related symptoms.

The app developed by AI FOR PET has been approved by the South Korean government as an animal medical device software. The AI ​​used in the app is based on an article written by South Korean Konkuk University and others published in 2019 in the British journal Scientific Reports.

The article is about a diagnosis model that uses a so-called convolutional neural network – a deep learning algorithm suitable for image recognition. The diagnostic model made it possible to determine the severity of ulcers based on photos of corneas with an accuracy of more than 90%

These innovations have emerged as the number of people keeping pets in Japan increases, especially after the coronavirus pandemic forced many workers to spend more time at home.

Starting this month, it will be mandatory for dogs and cats sold in Japan to be microchipped. And as AI gradually permeates all aspects of people’s lives, it may be inevitable that more and more companies look for ways to extend such technology into man’s best friend.