FireNet uses drones that fly above the wildfire and constantly record at 20 frames a second. This evidence is sent back to the AI in real time which decodes the data and identifies exactly where the wildfire is using GPS signals from the drones. Additionally, any rogue fire that splits of from the main wildfire is flagged by the software to analysts.
To train the system, it was fed tens of thousands of frames of wildfire videos. Thanks to machine learning, it is now claimed to spot the edges of fires with a 92% accuracy.