Skip to main content

Task 2 (under construction) : Few-shot sound detection

Table of Contents

User scenario
#

The model benchmark produced in this task should provide evidence-based answers to the following user needs:

“ I have just recovered a new dataset, which has been deployed specifically to record a targeted species A. I took time to annotate the first audio files but I don’t have time to go through the entire dataset. Also, quickly looking randomly at some other sound files, I realize that there might also be other species.

I already have a model for my target species A, but now what kind of models could I use to help me annotate the rest of my dataset ? Should I invest more effort in annotation before automatic tools can help me ? ”