F.L.A.R.E (Fire Localization and Response Engine) fuses computer vision, environmental sensors, and automation to spot fires early, pinpoint their location and type, and trigger the right suppression response โ all monitored live from one dashboard.
Four coordinated stages keep a space monitored around the clock and ready to react the instant something's wrong.
MQ-2 smoke and MQ-135 gas sensors continuously sample the air alongside the pod's camera.
A YOLOv8n detector scans every frame, and sensor fusion confirms the fire to rule out false alarms.
A vision-language model identifies the fire class while the arm aims at the fire's position in the frame.
Water or dry powder is dispensed automatically, an SMS goes out, and the event is logged to the dashboard.
Built for fast decisions during an emergency โ clear status, instant alerts, and history at a glance.
Pi and sensor-link status plus a rolling MQ-2 / MQ-135 gas chart, updated live.
Every detection event is time-stamped and archived with a snapshot and the AI's assessment.
The moment an anomaly is confirmed, an alert banner and notification fire off immediately.
A GSM module texts a fire alert straight to your phone, even when the internet is down.
Once the fire class is confirmed, the arm aims and releases water (Class A) or dry powder (Class B/C).
A capstone project from the University of the Cordilleras, College of Engineering and Architecture.
Open the dashboard to see live monitoring, alerts, and history.
Open Dashboard