Using Machine Vision for Detecting Wildfires

About 4.5 million U.S. homes are considered at high or extreme risk of wildfire. While wildfires mainly destroy unpopulated land and remain relatively “manageable” (with fewer than two percent classified as “significant”), wildfires that reach populated areas wreak havoc on those who live there. Furthermore, factors such as weather and geography make it difficult to predict fire movement and size. These limitations amplify the impact of unpredictable fires.

First responders need accurate and timely information to help combat the destruction wildfires cause to homes and businesses. Until recently, a lack of technology and infrastructure limited on-the-scene witnesses’ ability to act as an early warning system for an encroaching fire. While the recent advent of smartphones improves people’s ability to call or report a fire, networks often overload during emergencies. Furthermore, no efficient aggregator exists for the reported information, making it hard for emergency responders to judge potential threats.

Read about how we helped a technology company leverage Machine Vision to automate wildfire detection using a crowdsourced warning system designed to help save lives and property. The solution we partnered to build can identify wildfires with 98% accuracy using cameras on mobile devices.