Another use of computer vision/deep learning - IDing individual bears

There were several news articles this week profiling a project to ID individual brown bears in British Columbia and Alaska. The open access research article is available here: Automated facial recognition for wildlife that lack unique markings: A deep learning approach for brown bears

This Vancouver Sun article on the study also notes that UBC’s Marine Mammal Research Unit is focused on creating a similar system to identify orcas. Happywhale is doing something similar with humpback flukes, and they use photos from some iNat contributors to add to their data (here is my single contribution to that project).

And this is the link to a NYT article on the bear research, but you will need to create a free account to read it. An interesting link in the story is to, a site described as “an online broker of collaborations between technologists and conservationists”.


I saw that, about 84% accuracy rate, and they mentioned the ones it missed were largely ones researchers had trouble with also.

I wonder how many photos/angles of each individual bear are needed to reliably ID them, or if this only works for bears that already have dozens of photos of them?

Could be useful as a low-cost tool for population estimates I imagine.

1 Like

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.