Early Forest Fire Detection
Democratize open and low-tech solutions for fighting wildfires, for the benefit of the ecosystems and the citizens.
Temperate forests grow between the tropics and the boreal north, at latitudes of 25 to 50 degrees in both hemispheres, and cover 25% of the world’s forest area.1 Oak, beech, maple and birch share them with pine and fir.1
Why they matter
Much of Europe’s temperate forest was cleared long ago for farms and cities.1 What remains wild is precious: Romania still holds 250,000 hectares of virgin forest, most of it in the Carpathians, and the Southern Carpathians form Europe’s largest continuous unfragmented forest.2
Under pressure
Logging, mining, hunting, pollution and habitat loss all weigh on these forests,1 and in dry summers, so does wildfire.
What we do
With Pyronear, cameras on hilltops watch the forest and flag smoke early, so firefighters can reach a fire while it is still small. In the Carpathians, our bear deterrence work keeps bears and farmers apart at the forest edge.
Projects
Bear Identification
Noninvasive technologies to identify and monitor bears, facilitating their conservation.
Bear Deterrence in the Carpathians
Utilizing low-power technology to detect and deter bears from encroaching on Romanian farms.
Live demos
Temporal Smoke Verification
Watching camera sequences over time to tell genuine wildfire smoke from look-alikes.
Early Forest Fire Detection
Real-time camera analysis that spots wildfires early and raises the alarm fast.
Bear Identification
Facial recognition that tells individual bears apart from photos, tracking British Columbia's black bear population over time.
Bear Detector for Human Wildlife conflicts
Detecting and deterring bears near Romanian farms so people and predators can coexist.
Tools
Pyronear
Democratizing open and low-tech solutions for fighting wildfires. An early detection solution that is open source, efficient, automatic, energy-efficient, economical and modular.
Blog posts
Smoke Is a Behavior: Inside Pyronear's Temporal Wildfire Detection Model
How the temporal smoke verifier works, step by step. Detection boxes become tubes, stabilized crops make motion legible, and a vision transformer plus a tiny temporal head learn that smoke grows and drifts.
Racing Models, Not Opinions: How We Ran Wildfire ML R&D for Pyronear
How a literature survey, self-contained experiments, and a shared leaderboard turned 28 research papers into a production smoke verifier with 4x fewer false alarms.
Protecting the Forest: Building an early forest fire detector
Detecting early forest fires in real time using low powered technology
Open to collaboration
Every project here was built hand in hand with conservation partners, and much of the code is open source. If our work fits a challenge you're facing, or you'd like to build something together, we'd love to hear from you.