AI Wildfire Drones Advance From RGB Temperature Estimation to Autonomous Suppression
Clemson University researchers demonstrate an AI system that estimates wildfire temperatures from standard camera images, while Dryad Networks tests an autonomous drone that detected and extinguished a fire in under 12 minutes with no human intervention.
Two research efforts on opposite sides of the Atlantic are pushing drone-based wildfire response in complementary directions: one makes monitoring cheaper by replacing expensive thermal cameras with ordinary digital sensors, while the other eliminates human operators entirely by linking ground-based gas sensors to autonomous suppression drones.
Seeing Heat Through Color
Researchers at Clemson University have developed an AI system called SAM-TIFF that estimates wildfire temperatures from standard red-green-blue (RGB) photographs — the kind captured by any consumer-grade digital camera. The approach could reduce reliance on the thermal-imaging equipment that monitoring drones typically carry, which tends to be expensive, heavy, and power-hungry.
The team, led by associate professor of electrical and computer engineering Fatemeh Afghah, trained its models on pairs of RGB and radiometric thermal images collected during actual prescribed burns. By learning how subtle visual cues in color images correlate with heat signatures recorded in calibrated thermal data, the system can infer temperature maps using only an RGB camera feed during deployment.
The underlying research, described in a paper titled “Seeing Heat with Color,” won second place in the Best Student Paper contest at the IEEE Signal Processing Society’s Asilomar Conference on Signals, Systems, and Computers. While the system’s accuracy is lower than that of professional thermal cameras, which achieve precision within a few degrees, the Clemson team confirmed that the results remain strong enough to be useful when thermal equipment is unavailable or impractical.
The practical implications extend beyond cost savings. RGB-only temperature estimation enables lighter drone payloads and longer flight times, and it allows fire managers to identify hidden hotspots beneath smoke, ash, or vegetation in situations where thermal cameras may not be deployable. The research was supported by NASA and the National Science Foundation.
From Detection to Suppression in Under 12 Minutes
Where the Clemson project improves monitoring, German company Dryad Networks is attempting to close the entire loop from detection to extinguishment without human involvement. In a public demonstration held in Eberswalde, Germany, with XPRIZE Wildfire jury representatives present, the company showed its Silvaguard drone system autonomously responding to a controlled fire detected by its Silvanet ground sensor network.
The Silvanet system uses AI-powered gas sensors to identify fires by detecting volatile organic compounds released in the earliest stages of combustion. In the demonstration, the sensors flagged the fire within three minutes of ignition. A Silvaguard observation drone then launched autonomously from a solar-powered hangar, located the fire using optical and infrared imaging, and a suppression drone extinguished it — all within 12 minutes of ignition and with no human intervention.
“We have created the world’s first end-to-end autonomous wildfire suppression system combining ultra-early detection with AI-powered gas sensors, and autonomous observation and suppression drones,” said Dryad CEO Carsten Brinkschulte in a statement accompanying the demonstration.
Future iterations of Silvaguard are planned to include acoustic wave fire suppression technology, a method that uses directed sound waves to disrupt combustion. Dryad is a semifinalist in the XPRIZE Wildfire global competition, and the Silvaguard program has received partial funding through a 3.8-million-euro grant from the European Union’s European Regional Development Fund. The company’s Silvanet sensor network has already been deployed across 50 installations globally.
Complementary Approaches to a Growing Problem
The two projects address different points in the wildfire response timeline. Clemson’s RGB temperature estimation targets the monitoring phase, making it feasible to deploy cheaper and lighter drone fleets for continuous surveillance over fire-prone regions. Dryad’s Silvaguard targets the response phase, compressing the time between detection and suppression to minutes rather than the hours that traditional ground crews often require to reach remote ignitions.
Neither system has reached full operational deployment. The Clemson AI tool remains a research prototype that trades some accuracy for accessibility, and Dryad’s demonstration, while end-to-end, involved a controlled fire rather than a wildland scenario. Still, both efforts signal that drone-based wildfire technology is moving beyond passive observation toward active, autonomous intervention — a shift that could prove consequential as fire seasons lengthen and suppression resources thin.