Rainforest deforestation caused by illegal logging is a real threat to controlling rising CO2 levels. The quicker an alert is being able to be triggered, the more likely an intervention can be made.
Rainforest Connection repurposes old cellphones to monitor for illegal logging in the rainforest: "The moment a chainsaw goes off, our mic in the trees picks it up and we can then alert local rangers to stop people in the act... Up until now, satellites have been used to monitor for signs of deforestation but this doesn’t happen fast enough to halt the destruction. ''It is usually after the crime has been committed and there are more dangerous people on the ground. This real-time response has been helpful to people in the field.''"
The system "was built using TensorFlow, Google’s powerful open source computing resource. ''We can stream all of the audio in real time into the cloud, and even 20km from the nearest cellphone tower we can use TensorFlow to pick out the sounds of chainsaws.""
Audio files recovered from the rainforest
Today, according to Rainforest Connection’s website, the system monitors 26,000 hectares of forest, has gathered 4,629 days worth of data and has helped to sequester more than 6.5m metric tonnes of CO2, equal to taking 1.3m cars off the road. While the big-data problems are tantalising, from a machine-learning sense, making the network scalable and useful for rangers to react is a bonus in itself.