NEW YORK, Sept. 18 2025 — Fire kills nearly 3700 Americans annually and destroys $23 billion in property. Many deaths occur because traditional smoke detectors fail to alert occupants in time.
In response, researchers from NYU Tandon School of Engineering developed a technology that, unlike conventional smoke detectors that require significant smoke buildup and proximity to activate, spots fires in their earliest stages from video alone. The team's method operates within a cloud-based Internet of Things (IoT) architecture, where multiple standard security cameras stream raw video to servers that perform AI analysis.
The system analyzes video footage and identifies fires within 0.016 s per frame — faster than the blink of an eye— potentially providing crucial extra minutes for evacuation and emergency response.
"The key advantage [to the developed system] is speed and coverage," said lead researcher Prabodh Panindre, a research associate professor in NYU Tandon’s Department of Mechanical and Aerospace Engineering. "A single camera can monitor a much larger area than traditional detectors, and we can spot fires in the initial stages before they generate enough smoke to trigger conventional systems."
The need for improved fire detection technology is evident from concerning statistics: Eleven percent of residential fire fatalities occur in homes where smoke detectors failed to alert occupant either due to malfunction or the complete absence of detectors. Moreover, modern building materials and open floor plans have made fires spread faster than ever before, with structural collapse times significantly reduced compared to legacy construction.
The AI system analyzes raw video footage (left column) to detect fires and smoke using an AI ensemble. Blue boxes show detected fire, and red boxes show detected smoke. The system requires multiple algorithms to agree before confirming a fire detection. Courtesy of NYU.
The NYU Tandon research team developed an ensemble approach that combines multiple advanced AI algorithms. Rather than relying on a single AI model that might mistake a red car or sunset for fire, the system requires agreement between multiple algorithms before confirming a fire detection. This serves to substantially reduce false alarms.
The researchers trained their models by building a comprehensive custom image dataset representing all five classes of fires recognized by the National Fire Protection Association, from ordinary combustible materials to electrical fires and cooking-related incidents. It incorporates temporal analysis to differentiate between actual fires and static fire-like objects that could trigger false alarms. By monitoring how the size and shape of detected fire regions change over consecutive video frames, the algorithm can distinguish between a real, growing fire and a static image of flames hanging on a wall.
In tests, the system achieved high accuracy rates, with the best-performing model combination detection accuracy of greater than 80%.
"Real fires are dynamic, growing and changing shape," said Professor Sunil Kumar. "Our system tracks these changes over time, achieving 92.6% accuracy in eliminating false detections."
Upon the detection of fire, the system automatically generates video clips and sends real-time alerts via email and text message. This design means the technology can be implemented using existing CCTV infrastructure without requiring expensive hardware upgrades. Further, the technology can be integrated into drones or unmanned aerial vehicles to search for wildfires in remote forested areas. Early-stage wildfire detection would buy critical hours in the race to contain and extinguish them, enabling faster dispatch of resources, and prioritized evacuation orders that dramatically reduce ecological and property loss. And, to improve the safety of firefighters and assist during fire response, the same detection system can be embedded into the tools firefighters already carry: helmet cameras, thermal imagers, and vehicle-mounted cameras, as well as into autonomous firefighting robots. In urban areas, unmanned aerial vehicles integrated with this technology could help the fire service in performing 360-degree size-up, especially when fire is on higher floors of high-rise structures.
“It can remotely assist us in confirming the location of the fire and possibility of trapped occupants,” said Captain John Ceriello from the Fire Department of New York City.
Beyond fire detection, the researchers said that their approach could be adapted for other emergency scenarios such as security threats or medical emergencies, potentially expanding how we monitor and respond to various safety risks in our society.
The research was published in IEEE Internet of Things Journal (www.doi.org/10.1109/JIOT.2025.3598979).