Animal reID

Identify Every Individual

Non-invasive computer vision that tells one animal from another — bear, trout, seal, or snow leopard.

4
species supported
0 tags
fully non-invasive
100%
open source

Animal reID is a modular computer vision framework for identifying individual animals. It adapts to each species by choosing the right technique — facial recognition, spot-pattern matching, or local feature analysis.

From bears in British Columbia and trout in river systems to snow leopards in Central Asia and seals in coastal waters, it gives researchers non-invasive tools for wildlife monitoring and conservation.

See it in action

Upload a photo to the live demos and watch Animal reID pick out the individual — bear, trout, seal, or snow leopard.

Try the demos


Why Animal reID

Animal reID combines proven techniques into one adaptable system for individual identification.

One framework, many species

Proven on bears, trout, seals, and snow leopards, and extensible to new species.

The right technique for your data

Metric learning, local feature matching, or a hybrid, matched to your species and imagery.

Non-invasive by design

Identify individuals from camera-trap and observation images, with no tagging or marking.

Population trends over time

Match new sightings against historical databases to track survival, movement, and behavior.

Built to scale in the field

Automate identification to cut field costs and cover larger areas with existing networks.

Research-grade and open source

Built on peer-reviewed methods, with open-source implementations you can run and adapt.



Interactive Demos

Experience Animal reID in action with our live demonstrations. These systems are currently monitoring real wildlife populations.

Upload bear photographs and watch as the system segments facial features and matches them against our database of known individuals from British Columbia.


How It Works

Animal reID matches the identification technique to the species. Two approaches cover most cases:

Facial recognition

Metric Learning

Used successfully for bear identification, this approach combines instance segmentation to isolate animal faces with deep metric learning to create unique embeddings for each individual. The system learns to recognize subtle facial features and marking patterns that distinguish one animal from another.

Applications: bears, primates, big cats, and other species with distinctive facial characteristics.
  • Highly accurate for species with distinctive facial features
  • Robust to pose variations and lighting conditions
  • Proven in production for British Columbia bear monitoring
Spot patterns

Local Feature Matching

Pioneered in our trout identification work, this technique uses advanced local feature matching (LightGLUE) to analyze unique spot patterns on fish bodies. The system standardizes fish orientations and matches keypoint patterns against a reference database.

Applications: trout, leopards, cheetahs, whale sharks, and other spot-patterned species.
  • Non-invasive identification from natural markings
  • Works with partial views and occlusions
  • Effective for species with complex, unique pattern distributions


Resources & Documentation

Explore the open-source projects and technical guides behind Animal reID.

Each implementation is open-source with detailed documentation:

Bear Identification
Trout Identification
Snow Leopard Monitoring
Wadden Sea Seal Monitoring

Have a species to identify?

Tell us about your animals and your image data — we'll give you an honest read on which identification approach fits and what it would take.

Start a project