In Fontainebleau, forest surveillance is entering a new era. Installed on tall pylons, the Pyronear artificial intelligence system scans images captured by cameras every 30 seconds. If a suspicious plume appears, the alert is sent directly to the Seine-et-Marne fire department, who can then zoom in and locate the source of the fire. Energy-efficient and housed in a simple microcomputer, the technology has already proven itself this summer. When a disturbing odor was reported by walkers, the system made it possible to precisely triangulate the area of the fire. On July 14, it also helped assess the severity of a fire caused by a stolen car, mobilizing more resources than expected.
A promising system despite some failures
The Fontainebleau forest, once monitored by students stationed in wooden towers, now benefits from a permanent watch. The system isn't infallible: a harvester kicking up dust once triggered a false alarm. But according to firefighters, the AI learns from each mistake and becomes more efficient thanks to global databases and local retraining. One of its major assets remains its fuel-saving nature. No energy-hungry data centers: the algorithm runs locally and only trains for a few hours every two months. Easy to install on existing infrastructure, this system should be fully operational by next summer. Combining human vigilance and embedded technology, Fontainebleau is experimenting with a new way to protect its 25,000 hectares of oak trees and rocks. With a clear objective: to spot smoke as quickly as possible before it becomes fire.
