Get the latest tech news

Using squeezed wav2vec2 to automatically detect owl calls


How to use modern machine learning tools to recognise animal calls

I hope this blog post convinces you of two things: first, that near-state-of-the-art machine learning is easy to use even if you’re not a professional programmer, and second, that with a day or two of effort you can set up a pipeline that will reliably recognize any animal call (e.g. a birdcall), allowing you to record more data in more places. The genesis of this idea came from my partner, who came back from a field naturalist talk having been told that it was currently impossible to build an automatic recogniser for the call of the Australian Powerful Owl, Ninox strenua. I’m going to try to explain how to do this, assuming an audience that is not a professional programmer but has dabbled a bit with Python, and who has access to a reasonable volume of animal audio (say, a couple of hours).

Get the Android app

Or read this on Hacker News

Read more on:

Photo of wav2vec2

wav2vec2

Photo of owl calls

owl calls