AI Could Help Identify Individual Brown Bears in the Pyrenees

The debate over the brown bear in the Pyrenees has long focused on population figures, livestock losses, range expansion and coexistence with mountain farming. A new element is now entering that discussion: the possibility of identifying individual bears through facial recognition applied to images and video recorded in the wild.

According to Le Chasseur Français, brown bears in the Pyrenees could soon be identified from facial traits using a system inspired by the BearID project, an international initiative that applies artificial intelligence to individual bear recognition.

From counting bears to identifying individuals

The distinction matters. Traditional bear monitoring relies on signs of presence, camera traps, sightings, tracks and genetic analysis. Automated facial identification would not necessarily replace those methods, but it could complement fieldwork where enough high-quality images are available.

The BearID project presents itself as a machine-learning tool designed to identify individual bears from images and videos. Its aim is to support new non-invasive census and monitoring techniques for wild bears, particularly through remote cameras. The scientific basis of the system has been described in research on automated identification of brown bears using facial images.

For a species without obvious markings such as stripes or spots, identifying individuals can be especially difficult for non-specialists. AI systems look for stable patterns in the animal’s face, including the geometry between the eyes, muzzle and other facial reference points. Their usefulness will depend on image quality, model training and validation against established monitoring methods.

A sensitive Pyrenean context

The development comes at a time when brown bear monitoring in the Pyrenean range already depends on a cross-border technical framework. The Réseau Ours Brun, coordinated in France by the Office français de la biodiversité, recorded 3,287 indirect signs in 2025, of which 801 were submitted for genetic analysis, according to the annual report published in 2026.

Another report by the Office français de la biodiversité on damage assessments between 2019 and 2025 states that, in confirmed predation cases in France, the individuals involved were identified in 17.3% of cases. The same document warns of sampling bias and of the difficulty of interpreting an individual bear’s predatory behaviour solely from genetic detections.