New data challenge: “Few-shot Bioacoustic Event Detection” at DCASE 2021

[Dan Stowell <>]

Dear all,

Today we are announcing a new challenge in sound detection: "Few-shot
bioacoustic event detection", just launched as a part of the DCASE 2021
challenges.

    ** The core idea: given only 5 examples of an animal call, can a
system reliably detect that call type? **

Thanks to many collaborators, we publish new datasets, source code for 2
baseline systems (template matching and deep learning), and full
information at:

http://dcase.community/challenge2021/task-few-shot-bioacoustic-event-detection


Dates:

* 1st March: tasks and training data released
* 1st June: evaluation data released
* 15th June: deadline for submissions
* 1st July: challenge results

Please forward this to your students, colleagues - and anyone who might
be able to help in the mission to create reliable detection of animal
sounds!

Best wishes,
Dan Stowell
on behalf of the organising team


--
Dan Stowell
Associate Professor of AI & Biodiversity
* Tilburg University https://csai.nl/
* Naturalis Biodiversity Centre http://naturalis.nl/
* Jheronimus Academy of Data Sciences https://www.jads.nl/
* Queen Mary University of London http://machine-listening.eecs.qmul.ac.uk/

http://www.mcld.co.uk/research/