Two students came together with a shared aim of creating a permanent solution to wildfire prediction and prevention.
OK, so while news for big-tech in recent times has turned them from heroes into super villains, it’s important to remember that they do some good as well, and two students in California have utilized Google’s open source TensorFlow machine learning tool, to do just that.
Aditya Shah and Sanjana Shah, (unrelated) came together last year with a shared aim of creating a permanent solution to wildfire prediction and prevention.
Using TensorFlow, the two friends worked together to develop a device that could identify and predict areas in a forest that are susceptible to wildfires, providing an early warning to fire departments.
How it works
The device they came up with measures all the necessary ingredients needed to help predict the likelihood of forest fires. Factors such as wind speed, humidity, wind direction, temperature, and crucially even Biomass.
Biomass however, was the sticking point. It’s very hard to predict and measure things like falling branches, leaf fall, and dry scrub, and their consequent susceptibility to ignite.
Rather than giving up though, Shah and Shah, instead turned to Google’s machine learning AI TensorFlow to solve the problem.
TensorFlow was used to analyze images of biomass, estimate their size, moisture content, and determine the amount of ignitable foliage in a certain area.
Put simply, TensorFlow is being used to identify if biomass has a zero moisture content, and therefore likely to burst into flame, given the right conditions.
What it means for the future
The Shah’s then set up a network of sensors to link their Smart Wildlife Sensors together so that Fire Prevention crews can better predict where wild fires could start.
Giant sequoias, which are native to California are the world’s largest single trees and largest living thing by volume. In 2017, many thousands of trees were destroyed by wildfires in what was the most destructive wildfire season on record. More than 9,000 fires burning approximately 2,200 square miles of forest.
As Shah and Shah continue to improve their Smart Wildlife Sensor, it could mean that there are less fires in the future and that a truly beneficial and demonstrable use of Machine Learning has been found.