Brown Bears Look Alike to the Human Eye — An AI Program Is Helping to Observe Their Differences
Most brown bears may look like carbon copies of each other, but they all have their own little differences in appearance and behavior. Yet distinguishing bears can be a real challenge that requires a trained eye and years of careful observation. Fortunately, artificial intelligence makes bear watching much easier.
A new study published in Current biology has unveiled an AI program that can identify individual brown bears in the wild based on distinct facial features and poses. This has already shown promise in Alaska, where brown bears are regularly observed to understand how the species moves and behaves.
Brown bear sighting in Alaska
Alaska is home to approximately 30,000 brown bears, representing more than 98% of the U.S. brown bear population. Bear viewing sites across the state are popular destinations from July to September, the best time to see bears eating and playing.
One of these sites, the McNeil River State Game Sanctuary, is home to the largest seasonal gathering of brown bears in the world. Nearly 150 bears gather there in the summer, attracted by the masses of chum salmon that return to spawn in the McNeil River.
AI program helps identify brown bears in Alaska.
(Image credit: Beth Rosenberg/CC BY-SA 4.0)
The summer arrival of brown bears isn’t just for entertainment; it also gives scientists the opportunity to observe individual bears and keep tabs on population dynamics and ecosystem health.
However, brown bears are not so easy to distinguish from each other. They do not have clear markings and so it takes a lot of time and effort to distinguish each bear. This gets even trickier as the seasons change, as bears emerge from hibernation in the spring when they are at their skinniest, then change size and appearance later in the year.
Learn more: Orange rivers in Alaska mean color-changing crisis, exposing fish to toxic metals
Collect faces and poses
Researchers wanted to find a more practical way to identify individual bears, which is how the new AI model, PoseSwin, was born. The system focuses on facial features that do not change much over time: muzzle shape, brow bone angle, ear location, while also including pose information (including frontal, profile, and tilted views).
“Our biological intuition was that head features combined with pose would be more reliable than body shape alone, which changes dramatically with weight gain,” study author Alexander Mathis, a professor at the École Polytechnique Fédérale de Lausanne, Switzerland, said in a statement. “The data proved us right: PoseSwin significantly outperformed models that used body images or ignored pose information.”
AI helps identify bears in Alaska.
(Image credit: Beth Rosenberg/CC BY-SA 4.0)
The researchers trained PoseSwin by repeatedly showing the AI algorithm three images: two of the same bear taken at different times and one of a different bear. The program learned to group photos of the same bear next to each other, pushing out photos of other bears.
Leave the bears alone
PoseSwin was tested with photos taken by visitors to Katmai National Park and Preserve, located approximately 60 km from the McNeil River. The algorithm recognized individual bears from these photos, showing where they moved in search of food.
“The technology could eventually be used to analyze the thousands of photos that visitors take each year and help build a map of how brown bears use this vast area. This helps us understand what they need, how their population dynamics work, and many other important ecological questions,” said study author Beth Rosenberg, a researcher at Alaska Pacific University who captured more than 72,000 photos of 109 different brown bears from 2017 to 2022.
With PoseSwin, researchers can track bears without being invasive. Bears sometimes avoid areas where people see them, causing them to lose fat and protein from crucial feeding sites, according to the Alaska Department of Fish and Game.
Researchers observed bears – each with its own name, such as “Rocky,” “Sloth” and “Not Ears” – traveling between McNeil River and Katmai. Knowing how these bears move, they say, could provide insight into how the local ecosystem works.
Learn more: Are animals like mountain lions and bears leaving Yellowstone National Park?
Article sources
Our Discovermagazine.com editors use peer-reviewed research and high-quality sources for our articles, and our editors review the articles for scientific accuracy and editorial standards. See the sources used below for this article:
