Pyronear is a complete, open-source fire-detection system. A computer-vision model runs on a low-power microcomputer wired to cameras on high vantage points, watching the forest for the first signs of smoke. When it detects a fire, it sends an alert to a supervision platform used by fire departments — efficient, automatic, energy-efficient, and modular by design.
The video below, filmed in the Forest of Fontainebleau, shows how the system works end to end: a firefighter walks through the full pipeline, from the cameras spotting the first signs of smoke to the alert reaching the fire department.
End-to-end demonstration of the Pyronear system in the Forest of Fontainebleau
Why Pyronear
Pyronear pairs cutting-edge detection with open, low-tech hardware that communities can deploy themselves — tap each to learn more.
Computer-vision models detect wildfire smoke in real time and keep improving as new field data comes in.
An intuitive web app surfaces recent alerts and streamlines response for fire departments.
Built in the open on affordable, off-the-shelf hardware that communities can deploy and maintain themselves.
Runs on low-power microcomputers at the edge, keeping running costs and energy use low.
Detection, alerting, and the supervision platform are separable pieces you can adapt to local needs.
How It Works
Pyronear catches fires in two stages: a fast detector proposes candidate smoke, then a temporal model confirms it before an alert goes out.
Single-frame detection
A computer-vision model runs on each camera frame in real time, proposing candidate regions wherever it sees something that looks like smoke. It runs on a low-power microcomputer at the camera site, so only alerts — not video — travel over the network.
- Real-time inference at the edge on low-power hardware
- Continuously retrained as new field data arrives
- Sends lightweight alerts, not raw footage
Temporal verification
Candidate detections are tracked over time and a temporal classifier decides whether they actually behave like smoke. Real wildfires get confirmed within minutes, while clouds, fog, and haze look-alikes are filtered out before anyone is alerted.
- Tracks candidates across a full image sequence
- Rejects clouds, fog, and haze that fool single frames
- Keeps fire departments focused on real events
See Pyronear in Action
The computer vision model detected a forest fire in Fontainebleau from a distance of 35 kilometers in real time, setting a new record for the Pyronear system. The video below shows a thin black smoke rising in the distance.
Real-time forest fire detection from 35 kilometers away in FontainebleauTwo interactive demos let you run the Pyronear models on real data, right from this website.
Explore the full Pyronear platform
See live deployments, the fire-management platform, and how to get involved on the Pyronear website.
Visit Pyronear