When it comes to conservation, tracking specific animals usually involves collars or tags, but when software developers Ed Miller and Mary Nguyen saw the fast growth of AI for human facial recognition, they thought, “why can’t this be used for conservation efforts too?”
This idea, combined with inspiration from an Alaskan webcam broadcasting brown bears from Katmai National Park, led to the development of BearID, a system that uses AI to track specific bears and monitor population health. Miller and Nguyen worked with biologists Melanie Clapham and Chris Darimont to match their tech knowledge with biological research and develop the program.
Using AI to differentiate bears is not an easy task. Unlike giraffes or zebras, they have no unique individual markings, so the researchers focused on eye, nose, and ear positioning to train an AI system to recognize individual bears. After training the system with 3,740 photographed bear faces, the system learned to recognize them on its own.
Now that BearID is up and running, the researchers are tracking specific grizzly bears in Knight Inlet, Canada with 84% accuracy. This use of AI for conservation offers a more accurate and comprehensive, as well as less invasive, method for monitoring bear populations and their behaviors. As the team continues to perfect BearID, we are excited to see more applications of this solution throughout Canada and beyond.