Overview
Vision-based fire detection system identifying flames and smoke in real-time video feeds with sub-2-second alert generation for industrial safety applications.
The Problem
Traditional smoke detectors are slow to react in large open industrial spaces and can't pinpoint fire location. Visual detection enables faster response and precise localization.
The Solution
Developed a dual-model system: CNN for flame detection and smoke classification. Deployed on edge devices (Raspberry Pi 4) for distributed monitoring across industrial facilities.
Results & Impact
<2s
Alert Generation
From detection to notification
94%
Detection Accuracy
Combined fire + smoke
3
Deployment Sites
Industrial test facilities
Key Impact
- Faster response time than traditional detectors
- Reduced false alarms by 60% vs smoke detectors
- Provides visual evidence for incident analysis
Technologies Used
PythonTensorFlowOpenCVRaspberry PiIoTMQTT