Fabio Frazao
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Fabio Frazao has a total of 36 co-authors. Here are the top co-authors by number of shared publications:
- Aaron N. Rice
Cornell University
2 shared publications Audrey Looby
Institute of Food and Agricultural Sciences, University of Florida
2 shared publications- Christine Erbe
Curtin University
2 shared publications - Craig A. Radford
University of Auckland
2 shared publications - Francis Juanes
University of Victoria
2 shared publications - Jenni A. Stanley
University of Waikato
2 shared publications
Latest Publications
- 4 October 2024
- Published by Springer Nature
- p. 2149-2173
A Global Library of Underwater Biological Sounds (GLUBS): An Online Platform with Multiple Passive Acoustic Monitoring Applications
- 4 October 2024
- Published by Springer Nature
- p. 2149-2173
A Global Library of Underwater Biological Sounds (GLUBS): An Online Platform with Multiple Passive Acoustic Monitoring Applications
- Open Access
- 1 March 2024
- Published by Acoustical Society of America (ASA) in The Journal of the Acoustical Society of America
- Vol. 155 (3_Suppleme) , A87
Using a sequence deep learning model to increase the acoustic context of a killer whale detector
Although Killer whales (Orcinus orca) produce many stereotypical vocalizations, their sounds can be difficult to identify in isolation. Experts often rely on acoustic context to accurately identify these animals acoustically. Automated detectors and classifiers, on the other hand, frequently rely on short clips that capture individual vocalizations, not leveraging information regarding previous sounds. We developed deep learning models that used 1-minute inputs containing from 0 to 50 calls, with the average clip having 18. We tested three artificial neural network architectures that used recurrent layers to take the sequence of acoustic events into account. As a baseline, we used a convolutional neural network that only took 3-s clips at a time, without considering sequences of events. Here, we present preliminary evaluations on a dataset containing 360 min with Southern Resident killer whale activity in the Salish Sea, and an equal amount of data without killer whale sounds. The best model used a combination of temporal convolutional layers and gated recurrent units to achieve a recall of 0.95 at the maximum precision of 0.98. The models will be applied to near real-time monitoring efforts and will be open-sourced in the future.
- 1 March 2024
- Published by Acoustical Society of America (ASA) in The Journal of the Acoustical Society of America
- Vol. 155 (3_Suppleme) , A87
Open AccessUsing a sequence deep learning model to increase the acoustic context of a killer whale detector
Although Killer whales (Orcinus orca) produce many stereotypical vocalizations, their sounds can be difficult to identify in isolation. Experts often rely on acoustic context to accurately identify these animals acoustically. Automated detectors and classifiers, on the other hand, frequently rely on short clips that capture individual vocalizations, not leveraging information regarding previous sounds. We developed deep learning models that used 1-minute inputs containing from 0 to 50 calls, with the average clip having 18. We tested three artificial neural network architectures that used recurrent layers to take the sequence of acoustic events into account. As a baseline, we used a convolutional neural network that only took 3-s clips at a time, without considering sequences of events. Here, we present preliminary evaluations on a dataset containing 360 min with Southern Resident killer whale activity in the Salish Sea, and an equal amount of data without killer whale sounds. The best model used a combination of temporal convolutional layers and gated recurrent units to achieve a recall of 0.95 at the maximum precision of 0.98. The models will be applied to near real-time monitoring efforts and will be open-sourced in the future.
- Open Access
- 1 March 2024
- Published by Acoustical Society of America (ASA) in The Journal of the Acoustical Society of America
- Vol. 155 (3_Suppleme) , A242
An open-source acoustic detector for beluga whales, with evaluations in the Western Canadian Arctic
Belugas (Delphinapterus leucas) face threats from various sources, including noise from oil and gas exploration, vessel traffic, and other human activities. Here we present the development of a deep learning-based acoustic detector to automatically detect the species, and measure its performance when applied to study sites in the western Canadian Arctic. We used over 20,000 individually annotated beluga vocalizations to train deep learning models in the binary task of classifying 3-second audio clips into containing beluga vocalizations or not. Approximately 7,000 annotated vocalizations were reserved for testing, and models were evaluated on their ability to correctly label audio clips of two lengths: 3 s and 60 s. The average F1 score (across 10 models) on 3 s clips was 0.82 with a standard deviation of 0.027, with the best model achieving 0.86. When applied to 60 s clips, the best model achieved an F1 score of 0.96. We used the trained classifier to build a detector that processes longer recordings and will make it available as an open-source tool.
- 1 March 2024
- Published by Acoustical Society of America (ASA) in The Journal of the Acoustical Society of America
- Vol. 155 (3_Suppleme) , A242
Open AccessAn open-source acoustic detector for beluga whales, with evaluations in the Western Canadian Arctic
Belugas (Delphinapterus leucas) face threats from various sources, including noise from oil and gas exploration, vessel traffic, and other human activities. Here we present the development of a deep learning-based acoustic detector to automatically detect the species, and measure its performance when applied to study sites in the western Canadian Arctic. We used over 20,000 individually annotated beluga vocalizations to train deep learning models in the binary task of classifying 3-second audio clips into containing beluga vocalizations or not. Approximately 7,000 annotated vocalizations were reserved for testing, and models were evaluated on their ability to correctly label audio clips of two lengths: 3 s and 60 s. The average F1 score (across 10 models) on 3 s clips was 0.82 with a standard deviation of 0.027, with the best model achieving 0.86. When applied to 60 s clips, the best model achieved an F1 score of 0.96. We used the trained classifier to build a detector that processes longer recordings and will make it available as an open-source tool.