Bear Identification
-
Image Selection
Choose an image from the examples provided below, or upload your own data.
-
Run the ML model
Click the 'Submit' button to initiate the machine learning model.
-
Visualize the results
The system will segment bear faces from the images and compare them against our database of bear faces from British Columbia. It then identifies the individual bear with the highest probability match.
Overview
Bear Identification uses facial recognition to tell individual bears apart from ordinary photos — no collars, tags, or handling required. Developed in collaboration with the BearID Project, the system segments a bear’s face from an image and compares it against a catalog of known individuals from British Columbia, returning the closest match.
Because a bear’s facial features stay recognizable over time, the same animal can be re-identified across seasons and camera trap locations — turning everyday wildlife photos into long-term population data.
Key Features
What powers the identification:
Face detection & segmentation
Isolates the bear's face from a busy background so matching focuses on the features that actually distinguish one animal from another.
Metric-learning embeddings
Turns each face into a compact signature that captures the subtle traits separating individuals — the approach proven in our British Columbia work.
Catalog matching
Compares the query face against a database of known bears and ranks the most likely individuals by probability.
Confidence-ranked results
Surfaces the highest-probability match first, so reviewers can confirm an identity at a glance instead of scanning the whole catalog.
Fully non-invasive
Works from camera-trap and field photographs, with no tagging, collaring, or handling of the animals.
Use Cases
Where it supports bear conservation:
Population monitoring
Estimate how many distinct individuals use an area from camera-trap surveys, and track how that changes over time without physical capture.
Tracking individuals over time
Follow specific bears across seasons and camera locations to study survival, movement, and behavior.
Scalable field research
Automate identification that would otherwise take experts many hours of manual photo comparison, freeing time for analysis and fieldwork.
Learn more about the project
See the full bear identification project and our collaboration with the BearID Project — the field context and research behind this system.
View the project