At just 22, Oscar Mower built a machine learning model that can listen to raw ocean audio and pick out the call of a minke whale in real time, one of the most elusive and under-studied baleen whales in the world.
In this episode, Oscar explains how he turned whale song into images a computer could learn to recognise, why minke whales have been so hard to track using traditional methods, and how the same technology could one day help prevent ship strikes, one of the leading causes of whale mortality worldwide.
We also talk about the ongoing tension between abundant ocean audio data and the much scarcer resource of labelled data, why NOAA's open data repository made this project possible, and what a University of Sydney undergraduate research paper turned into: a tool with the potential to help map minke whale migration routes for the first time.
More Information
https://www.fisheries.noaa.gov/national/science-data/sounds-ocean-environmental-and-anthropogenic
If you enjoy this podcast, please like and subscribe to our show wherever you get your podcasts.
Leave us a review and share this show with your friends.
It really helps us to reach more citizen scientists, like you.
Contact the Show
We are always looking for more guests to tell us about interesting citizen science projects, research and events.
You can email us at: info@citizenscienceshow.com