This post is also available in:
עברית (Hebrew)
Researchers at NYU Tandon School of Engineering have developed an AI system capable of detecting fires much earlier than traditional smoke detectors, potentially saving lives and property. By utilizing regular security cameras, this system analyzes video in real-time to identify smoke or fire in its initial stages.
Unlike conventional smoke detectors, which rely on significant smoke buildup to trigger alarms, this AI model processes video feeds at a speed of 0.016 seconds per frame—far faster than human reaction time. This allows the system to detect fires in their earliest moments, providing vital extra minutes for evacuations and emergency responses, according to TechXplore.
Published in the IEEE Internet of Things, one major advantage of the technology is its expansive coverage. While traditional smoke detectors are limited to small areas, a single camera can monitor much larger spaces, helping to identify fires before they have spread. The system uses a combination of multiple AI algorithms to analyze the footage, ensuring more accurate detections and reducing the likelihood of false alarms, which can occur when non-fire events are mistakenly identified as fires.
To improve accuracy, the system has been trained on a comprehensive dataset representing various types of fires. This results in a detection accuracy of 80.6% in optimal conditions, with the ability to achieve 92.6% accuracy in eliminating false positives by analyzing how fire patterns change over time in video frames.
The system operates within a cloud-based infrastructure, enabling integration with existing CCTV networks without requiring major hardware upgrades. Once a fire is detected, the system immediately sends alerts through text and email, along with video clips, allowing for a rapid response.
This technology also has the potential to be adapted for wildfire detection using drones, or to support firefighting teams with real-time data from helmet cameras and thermal sensors. The flexibility of this AI-driven solution could revolutionize fire safety across multiple sectors, improving both early detection and emergency management.
