48 Hours Of Pollution And Littering Around The World Through Deep Learning

We've only just begun to explore the incredible potential in using the Google Cloud Vision API to catalog the world's news imagery, yet one early application that struck us was the ability to use the API to generate a better understanding of the state of our planet, especially society's impact on our natural world through pollution. Its one thing to read an abstract headline that there is air pollution in some city halfway across the globe and quite another to actually see that smog through the imagery of local media.

Towards this end, we ran a simple test using the Cloud Vision API to filter for all coverage monitored by GDELT from around the world over the last 48 hours containing imagery that the API identified as "smog" and a separate search (at the bottom of this post) where we searched for imagery identified by the API as either "trash" or "litter."

The results are profoundly striking, offering a sobering look at the state of our planet.  Imagine this collage updated every day as a realtime view onto the state of our world.

 

SMOG

Smog-Vision-API-Examples

 

LITTERING

Litter-Vision-API-Examples