This morning at 06:08 local time, the polar bear detector that Lars Holst Hansen and Kim Hendrikse from Wildlife Security Innovations have been working on for more than a year detected a polar bear in Greenland.
The system is scanning 24 video streams viewing each camera once every couple of seconds looking for polar bears. 7 of those streams are real-time digital zooms providing for reliable longer distance detection. All this is happening on just one Jetson edge computer.
When a bear is detected an alert is generated as well as waking a talking monitoring system that speaks where the bear was seen. Lars in Greenland receives alerts as well as me in the Netherlands. The talking monitoring system is just a simple raspberry with a USB speaker but can monitor as many remote systems as you configure.
This bear managed to only have a very quick look around Zackenberg research station before being scared off.
Scared off, saving the lives off both the bear and people.
With the successful detection this morning, I'm proud to say, the system has proven itself to be a working, practical and scalable solution.
23 August 2025 10:26am
Wow! It’s amazing to see the system and AI working great on real data. Congratulations
27 August 2025 7:01pm
Great work!
Does your bear detection model avaliable for 3rd party usage?
I`m working on a prototype project for tracking bears in a zoo enclosure. I`ve trained the yolo11n model on the bears dataset and would like to compare the results.
Hugo Markoff