Artificial Intelligence Uses Smartphone Photos From All Over the World to Track Pollution


What if managing city smog were as simple as snapping a photo? Maybe it is. A new project called AirTick is using cellphone snaps to help scientists track air pollution.

The project started with a fundamental question: How do we manage good air quality?

Photos taken with AirTick's signature mobile app are analyzed and mined for their meta data (where and when the photo was taken). That information is then checked against official state data about air quality to estimate air pollution. Using a learning machine called the Deep Boltzmann, AirTick's team hopes that with enough data, the app will be able to build an algorithm strong enough to estimate real-time air quality just using photos. 

China is notorious for air pollution, so much so, its citizens have been known to buy bottled Canadian air. In fact, much of Southeast Asia operates in a haze of pollution so disruptive, it's been known to close schools.  

Jedi Pan Zhengxiang, the research scholar behind AirTick, said putting up a bunch of air sensors is too costly a prospect and inefficient. Air sensors typically take up a lot of space and smaller air sensors are too expensive to scale. Plus who wants to walk around with an air sensor?

Getty Images

But most people in countries where smog is a major health concern have mobile phones. In India, a country with some of the worst air pollution in the world, more people have access to cellphones than they do to a toilet, according to the United Nations University

And the health risks of air pollution are not lost on the people living in these regions. Some people are leaving New Dehli over health concerns; still others have fixed their aim on the Indian government with a Supreme Court petition arguing that the levels of air pollution are unconstitutional, according to the Financial Times

What that means for AirTick is that people are interested in these issues and potentially willing to contribute to a crowdsourced project that would, at the very least, help people get a better understanding of just how bad pollution is on a given day.  

h/t New Scientist