Detectors react late
A smoke detector only triggers when smoke reaches it. Under high ceilings, in large halls and outdoors, that can take minutes — or never happen.
AI fire & smoke detection · Made in Algeria
See fire and smoke the moment they start.
ForestEye turns ordinary security cameras and low-cost sensors into an early fire alarm: AI spots flames and smoke within seconds, sounds the alarm on site and alerts your team wherever they are — before a spark becomes a disaster.
Developed at Mohamed El Bachir El Ibrahimi University, Bordj Bou Arréridj, and its business incubator.
Fire detected16:53
Camera 2 · confidence 58% · alarm sounding
Sensor node 01
41 °C ↑18 %
The problem
Most fires are seen too late
A fire is easiest to stop in its first minutes. Yet the usual ways of spotting it often react only when it has already grown.
Cameras nobody watches
Security cameras are everywhere, but no one can watch every screen 24 hours a day — least of all at night and on weekends.
Every minute counts
The earlier a fire is seen, the smaller it is, the safer people are and the less is lost. Minutes decide between an incident and a disaster.
Our solution
Every camera becomes a fire watcher
Electronic eyes that never sleep, sensors that feel the first signs of fire, and a platform that turns them into confirmed alerts.
AI fire & smoke detection
Camera streams are analysed in real time by a model trained on real fire and smoke images, with extra checks on colour, motion and persistence to cut false alarms.
Field sensors
A low-power ESP32 alarm box measures temperature, humidity, flame and gas — and sounds a local alarm even when the network is down.
Instant alerts
An alert reaches the on-site alarm in a fraction of a second and the live portal at the same time, with a snapshot and a video clip.
One live portal
Live video, alerts, sensor readings and history for every site, accessible securely from anywhere.
How it works
From the first spark to the right people
Detect
Cameras and sensors watch your site 24/7. The AI examines every frame for flames and smoke.
Confirm
A detection must persist and pass colour and motion checks before it becomes an alert.
Alert
The local alarm sounds, the team sees the alert in the portal with a snapshot, and a clip is recorded.
The platform
Built and tested on real hardware
Our working prototype links an IP camera, an ESP32 alarm and sensor box, and the ForestEye portal. These are measured results, not targets.
- median time from alert to on-site alarm
- 0.2 smedian time from alert to on-site alarm
- no GPU needed to run the AI
- CPU onlyno GPU needed to run the AI
- fire-model analysis per frame on a laptop CPU
- ≈ 25 msfire-model analysis per frame on a laptop CPU
- works with standard IP cameras
- RTSPworks with standard IP cameras
Who it's for
Wherever a fire must be caught early
Factories & warehouses
Large volumes and high ceilings where detectors react late, and stock worth protecting.
Offices & public buildings
Schools, administrations and offices watched around the clock, nights and weekends included.
Farms & storage
Barns, hay, grain and machinery, often far from anyone who could see a fire.
Car parks & outdoor sites
Open spaces where smoke detectors cannot work at all.
Homes & small businesses
An alert on your phone when nobody is there, at night or on holiday.
Forests & natural areas
Cameras on masts that watch the horizon for the first smoke of a wildfire.
Let's protect your sites together
Tell us about your site. We'll show you ForestEye working live, and plan a pilot with you.