Using Machine Learning to save the rainforest

Recently I attended a meetup titled “Lightning talks – Impact driven AI applications” at University of San Francisco. The main speakers of the event were renowned data scientists like Jeremy Howard and Rachel Thomas. The session focussed on discussing about the various applications of machine learning, right from ways of using word embeddings to improve NLP for Urdu and other low-resource languages (Talk by Samar Haider) to using deep learning to predict breast cancer at an early stage (Talk by Ljubomir Buturovic).

Machine Learning Rainforest Connection

There was one particular talk by Sara Hooker and Sean McPherson which really intrigued me. They discussed how they are working towards saving rainforest by use of machine learning. Never in my wildest dream had I ever imagined this particular application of machine learning. Sara Hooker is the Executive Director of Delta Analytics that works with non-profits who use machine learning for the social good. One particular project Delta had been associated with is Rainforest Connection.

As per Delta Analytics, they are working with Rainforest Connection “to improve and build upon existing models that use audio data streamed from recycled cell phone sensors positioned in endangered forests. The goal is to track ecological markers such as the presence of specific species, and harmful human activity such as illegal logging. Rainforest Connection wishes to combine existing models to improve accuracy, and add new models that take into account more information, such as the distance between a sensor and an event in the forest. Improved models will enable Rainforest Connection to better serve scientific and environmental partners by providing more precise and accurate real-time information about the presence of certain species and the overall health of an environment.”

In the session, Sara along with Sean discussed how they are using simple machine learning techniques to reduce illegal logging. The idea is to install audio sensors throughout the rainforests which listens to audio signals being emitted in their periphery. The audio signal is then processed to determine whether the signal is coming from a chainsaw. So basically, the model they have developed can differentiate the sound of a chainsaw from other sounds using data science. Further, they are also able to predict which direction is the sound being emitted from so that rangers can be immediately notified of suspected illegal logging.

Isn’t it a wonderful application of science and technology which has a direct impact on the environment? Please feel free to comment.

Spread the word!
  •  
  •  
  •  
  •  
  •  
  •  
  •  
  •  
  •  
  •  
  •  
  •  

Leave a Reply