Description
Polar bears pose a significant danger to people, and encountering one in the open can be life‑threatening. Lars Holst Hansen of Aarhus University, who periodically works at the Zackenberg research station in Greenland, envisioned an early‑warning system that could protect both researchers and polar bears. Field teams at the station must travel armed, and unexpected encounters carry a high risk of injury. Early detection reduces the likelihood of surprise encounters and helps avoid lethal outcomes. Founder Kim Hendrikse connected with Lars through the wildlife collaboration platform Wildlabs, leading to the creation of the project.
What we did
- Collected a large set of polar bear images from Copenhagen Zoo and Skandinavisk Dyrepark, using a pre‑trained YOLOv6‑large model
- Compiled a suitable set of alternative target images from open‑source archives
- Trained a high‑quality YOLOv9 model
- Deployed the system in Greenland on an NVIDIA Jetson platform
Results
- The system successfully detected polar bears in the wild in Greenland on two separate occasions
Youtube
Below is the Wildlabs presentation of the project
Want something similar?