Animal world Research team develops facial recognition program for bears
Facial recognition for bears: A Lausanne research team has developed an AI program that can distinguish individual brown bears from one another.
29.01.2026, 10:01
SDA
The program also works if the bears are emaciated after hibernation or have put on a lot of fat for the winter. As the Swiss Federal Institute of Technology in Lausanne (EPFL) announced on Thursday, the system is intended to facilitate wildlife research and species protection.
According to the university, visually distinguishing a particular brown bear from its conspecifics requires years of experience and a trained eye. To make matters worse, bears in Alaska come out of hibernation in spring with shaggy fur and significant weight loss, only to gain considerable weight when they catch salmon and lose their winter fur at the same time.
Head features and body posture are key
To develop their detection program, the researchers from EPFL and Anchorage-based Alaska Pacific University focused on features that change little over time: the shape of the snout, the angle of the frontal bones and the position of the ears. According to the researchers, the decisive factor is the combination of these features with the animals' posture.
The program was trained with over 72,000 photos of 109 different brown bears. The images were taken between 2017 and 2022 in a wildlife sanctuary in Alaska.
In an initial test, the AI was fed with photos taken by visitors from the nearby Katmai National Park. The program was able to recognize several bears from the protected area and thus detect their seasonal migrations in search of food.
The technology should also be applicable to other animal species and has already demonstrated a high level of accuracy in tests with macaques. The researchers have published the technology in the journal "Current Biology". The algorithm has also been published as open source so that other research groups can use it.
