Combining two user-friendly machine learning tools increases species detection from acoustic recordings
Please use this identifier to cite or link to this item:
http://hdl.handle.net/10045/139550
Title: | Combining two user-friendly machine learning tools increases species detection from acoustic recordings |
---|---|
Authors: | Pérez-Granados, Cristian | Feldman, Mariano J. | Mazerolle, Marc J. |
Research Group/s: | Ecología y Conservación de Poblaciones y Comunidades Animales (ECPCA) |
Center, Department or Service: | Universidad de Alicante. Departamento de Ecología |
Keywords: | American toad | BirdNET | Convolutional neural network | Kaleidoscope Pro | Anaxyrus americanus (Holbrook, 1836) | Passive acoustic monitoring |
Issue Date: | 2-Jan-2024 |
Publisher: | Canadian Science Publishing |
Citation: | Canadian Journal of Zoology. 2024, 102(4): 403-409. https://doi.org/10.1139/cjz-2023-0154 |
Abstract: | Passive acoustic monitoring usually generates large datasets that require machine learning algorithms to scan sound files, although the complexity of developing machine learning algorithms can be a barrier. We assessed the ability and speed of two user-friendly machine learning tools, Kaleidoscope Pro and BirdNET, for detecting the American toad (Anaxyrus americanus (Holbrook, 1836)) in sound recordings. We developed a two-step approach, combining both tools to maximize species detection while minimizing the time needed for output verification. When considered separately, Kaleidoscope Pro successfully detected the American toad in 85.9% of recordings in the validation dataset, while BirdNET detected the species in 58.4% of recordings. Combining the two tools in the two-step approach increased the detection rate to 93.3%. We applied the two-step approach to a large acoustic dataset (n = 6194 recordings). We started by scanning the dataset using Kaleidoscope Pro (species detected in 417 recordings), then we used BirdNET on the remaining recordings without confirmed presence. The two-step approach reduced the scanning time, the time needed for output verification, and added 37 additional species detections in 45 min. Our findings highlight that combining machine learning tools can improve species detectability while minimizing time and effort. |
Sponsor: | CP acknowledges the support of the Ministerio of Educación y Formación Profesional through the Beatriz Galindo Fellowship (Beatriz Galindo—Convocatoria 2020). MJF gratefully acknowledges the field assistance of R. Chevallier in setting up the acoustic equipment, and L. Imbeau and N. Fenton for their contributions to the conception and design of the study. The Natural Sciences and Engineering Research Council of Canada—UQAT Industrial Research Chair on Nordic Biodiversity in a Mining Context funded the field work. |
URI: | http://hdl.handle.net/10045/139550 |
ISSN: | 0008-4301 (Print) | 1480-3283 (Online) |
DOI: | 10.1139/cjz-2023-0154 |
Language: | eng |
Type: | info:eu-repo/semantics/article |
Rights: | © 2024 The Author(s). This work is licensed under a Creative Commons Attribution 4.0 International License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author(s) and source are credited. |
Peer Review: | si |
Publisher version: | https://doi.org/10.1139/cjz-2023-0154 |
Appears in Collections: | INV - ECPCA - Artículos de Revistas |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
Perez-Granados_etal_2024_CanJZool.pdf | 480,98 kB | Adobe PDF | Open Preview | |
Items in RUA are protected by copyright, with all rights reserved, unless otherwise indicated.