About the project
In collaboration with WWF Romania and Sensing Clues, we monitor brown bears in real time across the Romanian forests. The system pairs 4G cameras in the field with AI classification on the backend, so researchers and conservationists know within seconds which species turned up where — without anyone having to walk in and pull an SD card.
How it works
For this project we deployed the 4G variant of Instant Detect — cameras that capture an image on motion and push it directly over 4G to our backend. There, a two-stage AI pipeline takes over:
- MegaDetector — a first model that locates where in the image an animal is and crops out the relevant regions
- SpeciesNet — a second model that classifies each cropped detection: which species is this, exactly?
Splitting detection from classification keeps the pipeline robust. MegaDetector is trained broadly on everything camera traps catch; that lets SpeciesNet focus on accurately classifying specific species, instead of first having to filter out leaves, shadows or rain triggering the camera.
Sensing Clues integration
The results — species, location, timestamp and image context — are pushed into the Sensing Clues platform, where the WWF teams and researchers analyse the data, spot patterns and plan their field work. The information doesn’t get stuck in our backend; it lands directly in the tools the teams already work with every day.
Why it matters
Brown bears in Romania are a significant but vulnerable population. Knowing where they are — and when — is essential both for research into behaviour and migration, and for preventing human–wildlife conflict. Real-time data from the field dramatically sharpens that picture, and lets teams act quickly rather than analyse weeks-old captures after the fact.
