Few-shot bioacoustic event detection at the DCASE 2023 challenge

Ines Nolasco; Burooj Ghani; Shubhr Singh; Ester Vidaña-Vila; Helen Whitehead; Emily Grout; Michael Emmerson; Frants Jensen; Ivan Kiskin; Joe Morford; Ariana Strandburg-Peshkin; Lisa Gill; Hanna Pamuła; Vincent Lostanlen; Dan Stowell

Few-shot bioacoustic event detection at the DCASE 2023 challenge

Ines Nolasco, Burooj Ghani, Shubhr Singh, Ester Vidaña-Vila, Helen Whitehead, Emily Grout, Michael Emmerson, Frants Jensen, Ivan Kiskin, Joe Morford, Ariana Strandburg-Peshkin, Lisa Gill, Hanna Pamuła, Vincent Lostanlen, Dan Stowell

Research output: Contribution to conference › Paper › Scientific › peer-review

7 Downloads (Pure)

Abstract

Few-shot bioacoustic event detection consists in detecting sound events of specified types, in varying soundscapes, while having access to only a few examples of the class of interest. This task ran as part of the DCASE challenge for the third time this year with an evaluation set expanded to include new animal species, and a new rule: ensemble models were no longer allowed. The 2023 few shot task received submissions from 6 different teams with F-scores reaching as high as 63% on the evaluation set. Here we describe the task, focusing on describing the elements that differed from previous years. We also take a look back at past editions to describe how the task has evolved. Not only have the F-score results steadily improved (40% to 60% to 63%), but the type of systems proposed have also become more complex. Sound event detection systems are no longer simple variations of the baselines provided: multiple few-shot learning methodologies are still strong contenders for the task.

Original language	English
Number of pages	5
Publication status	Published - 15 Jun 2023
Event	Detection and Classification of Acoustic Scenes and Events 2023 - Tampere, Finland Duration: 21 Sept 2023 → 22 Sept 2023

Conference

Conference	Detection and Classification of Acoustic Scenes and Events 2023
Abbreviated title	DCASE 2023 workshop
Country/Territory	Finland
City	Tampere
Period	21/09/23 → 22/09/23

Keywords

cs.SD
cs.LG
eess.AS

Access to Document

https://arxiv.org/pdf/2306.09223.pdf

Cite this

Nolasco, I., Ghani, B., Singh, S., Vidaña-Vila, E., Whitehead, H., Grout, E., Emmerson, M., Jensen, F., Kiskin, I., Morford, J., Strandburg-Peshkin, A., Gill, L., Pamuła, H., Lostanlen, V., & Stowell, D. (2023). Few-shot bioacoustic event detection at the DCASE 2023 challenge. Paper presented at Detection and Classification of Acoustic Scenes and Events 2023, Tampere, Finland. https://arxiv.org/pdf/2306.09223.pdf

@conference{65cf9a731c9849e998cb604dff068138,

title = "Few-shot bioacoustic event detection at the DCASE 2023 challenge",

abstract = "Few-shot bioacoustic event detection consists in detecting sound events of specified types, in varying soundscapes, while having access to only a few examples of the class of interest. This task ran as part of the DCASE challenge for the third time this year with an evaluation set expanded to include new animal species, and a new rule: ensemble models were no longer allowed. The 2023 few shot task received submissions from 6 different teams with F-scores reaching as high as 63% on the evaluation set. Here we describe the task, focusing on describing the elements that differed from previous years. We also take a look back at past editions to describe how the task has evolved. Not only have the F-score results steadily improved (40% to 60% to 63%), but the type of systems proposed have also become more complex. Sound event detection systems are no longer simple variations of the baselines provided: multiple few-shot learning methodologies are still strong contenders for the task.",

keywords = "cs.SD, cs.LG, eess.AS",

author = "Ines Nolasco and Burooj Ghani and Shubhr Singh and Ester Vida{\~n}a-Vila and Helen Whitehead and Emily Grout and Michael Emmerson and Frants Jensen and Ivan Kiskin and Joe Morford and Ariana Strandburg-Peshkin and Lisa Gill and Hanna Pamu{\l}a and Vincent Lostanlen and Dan Stowell",

note = "submitted to DCASE 2023 workshop; Detection and Classification of Acoustic Scenes and Events 2023, DCASE 2023 workshop ; Conference date: 21-09-2023 Through 22-09-2023",

year = "2023",

month = jun,

day = "15",

language = "English",

}

Nolasco, I, Ghani, B, Singh, S, Vidaña-Vila, E, Whitehead, H, Grout, E, Emmerson, M, Jensen, F, Kiskin, I, Morford, J, Strandburg-Peshkin, A, Gill, L, Pamuła, H, Lostanlen, V & Stowell, D 2023, 'Few-shot bioacoustic event detection at the DCASE 2023 challenge', Paper presented at Detection and Classification of Acoustic Scenes and Events 2023, Tampere, Finland, 21/09/23 - 22/09/23. <https://arxiv.org/pdf/2306.09223.pdf>

TY - CONF

T1 - Few-shot bioacoustic event detection at the DCASE 2023 challenge

AU - Nolasco, Ines

AU - Ghani, Burooj

AU - Singh, Shubhr

AU - Vidaña-Vila, Ester

AU - Whitehead, Helen

AU - Grout, Emily

AU - Emmerson, Michael

AU - Jensen, Frants

AU - Kiskin, Ivan

AU - Morford, Joe

AU - Strandburg-Peshkin, Ariana

AU - Gill, Lisa

AU - Pamuła, Hanna

AU - Lostanlen, Vincent

AU - Stowell, Dan

N1 - submitted to DCASE 2023 workshop

PY - 2023/6/15

Y1 - 2023/6/15

N2 - Few-shot bioacoustic event detection consists in detecting sound events of specified types, in varying soundscapes, while having access to only a few examples of the class of interest. This task ran as part of the DCASE challenge for the third time this year with an evaluation set expanded to include new animal species, and a new rule: ensemble models were no longer allowed. The 2023 few shot task received submissions from 6 different teams with F-scores reaching as high as 63% on the evaluation set. Here we describe the task, focusing on describing the elements that differed from previous years. We also take a look back at past editions to describe how the task has evolved. Not only have the F-score results steadily improved (40% to 60% to 63%), but the type of systems proposed have also become more complex. Sound event detection systems are no longer simple variations of the baselines provided: multiple few-shot learning methodologies are still strong contenders for the task.

AB - Few-shot bioacoustic event detection consists in detecting sound events of specified types, in varying soundscapes, while having access to only a few examples of the class of interest. This task ran as part of the DCASE challenge for the third time this year with an evaluation set expanded to include new animal species, and a new rule: ensemble models were no longer allowed. The 2023 few shot task received submissions from 6 different teams with F-scores reaching as high as 63% on the evaluation set. Here we describe the task, focusing on describing the elements that differed from previous years. We also take a look back at past editions to describe how the task has evolved. Not only have the F-score results steadily improved (40% to 60% to 63%), but the type of systems proposed have also become more complex. Sound event detection systems are no longer simple variations of the baselines provided: multiple few-shot learning methodologies are still strong contenders for the task.

KW - cs.SD

KW - cs.LG

KW - eess.AS

M3 - Paper

T2 - Detection and Classification of Acoustic Scenes and Events 2023

Y2 - 21 September 2023 through 22 September 2023

ER -

Few-shot bioacoustic event detection at the DCASE 2023 challenge

Abstract

Conference

Keywords

Access to Document

Fingerprint

Cite this