Scientists are using AI facial recognition to identify bears

Researchers are using facial recognition to identify bears
Image: Melanie Clapham | BearID

Facial recognition can be found just about everywhere these days. From the iPhone’s FaceID to surveillance cameras that use the tech to spot criminals, algorithms that identify faces are becoming increasingly popular.

However, it isn’t just human faces that they are learning to spot. Researchers believe that facial recognition can be helpful in areas like conservation by using it to identify wild animals.

Scientists from British Columbia, Canada, are testing that theory with grizzly bears. After spending more than a decade studying the creatures, they are now using artificial intelligence (AI) to identify them. The team believes that facial recognition could be a more humane and cost-effective method of tracking animals to study their behaviors and migration patterns.

Manage your supply chain from home with Sourcengine


Melanie Clapham is a bear biologist. It’s a small field, but one that is extremely important to the conservation of the species. Several years ago, she teamed up with Silicon Valley tech employees to create a tool called BearID.

Its purpose is to identify individual grizzly bears in the wild. So far, the program is able to recognize 132 of the animals.

Interestingly, BearID was adapted from an existing piece of animal-focused AI software called Dog Hipsterizer. The latter is used to overlay virtual mustaches and hats atop photos of dogs. The researchers have adapted the underlying code so that it can be used to identify bears based on images from cameras in Alaska’s Katmai National Park.

The team uses unique features on the bears’ faces to differentiate them. Things like scars, cuts, and patches of coloration help the software determine if it has seen a particular bear before.

The researchers say that BearID has collected 4,674 images of grizzly bears. Of those photos, 80 percent have been used to train the AI while the other 20 percent are for testing purposes.

According to a study published in the journal Ecology and Evolution, BearID is now 84 percent accurate. That’s an impressive accomplishment considering that some human facial recognition programs struggle to earn such high marks.

Of course, the only downside is the fact that the bear in question must already be part of the dataset before it can be identified. As the project grows, that will be less of an issue. Perhaps the researchers will even find a way to add a bear as a new individual if it hasn’t been seen before.

Cows Want In, Too

It appears that identifying bears with AI isn’t the only animal application for the technology. Cattle ranchers believe it could also be beneficial for tracking domesticated animals.

Joe Hoagland, a rancher from Kansas, is building an app called CattleTracs. It allows anyone to take a photo of a cow and submit it to an online database along with its GPS coordinates and the date. Then, animals that are matched in the database can be tracked over time. This could be particularly useful when dealing with a herd illness or a runaway animal.

Hoagland says, “Being able to trace that diseased animal, find its source, quarantine it, do contact tracing—all the things we’re talking about with coronavirus are things we can do with animals, too.”

Facial recognition is a hotly-disputed topic when it is being used on humans. However, it seems that the tech could be beneficial—with fewer privacy concerns—in the animal world.

That being said, researchers fear that it could also be subverted by poachers. In the same way that it could be useful for tracking domesticated animals, it could lead malicious hunters to their prey with ease.

Still, it appears that projects like BearID and CattleTracs present more opportunities for good than harm at this point.


Please enter your comment!
Please enter your name here