EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Effects of network selection and acoustic environment on bounding-box object detection of delphinid whistles using a deep learning tool

Peter C Sugarman, Elizabeth L Ferguson iD, Gabriela C Alongi, Jennifer Pettis Schallert, Heidi Lyn

The Journal of the Acoustical Society of America · 2025

DOI: 10.1121/10.0036942 ↗

Kurz gefasst

Fachartikel in der Zeitschrift The Journal of the Acoustical Society of America, 2025. Verfasst von Peter C Sugarman und 4 weiteren Autor:innen. Beteiligte Einrichtung: Acoustic Interactions, LLC, Bellevue, Washington 98008, USA. Hauptthemen laut MEDLINE: Deep Learning, Vocalization, Animal, Bottle-Nosed Dolphin, Acoustics. Die Arbeit wurde bisher 5-mal zitiert (Crossref) und nennt 49 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Deep learning methods offer automated solutions for detecting marine mammal calls, yet require time-intensive development for optimized neural network performance, including carefully curating data and creating a robust network architecture. Using data collected in two aquarium and two open ocean environments, we evaluated the performance of a series of pre-trained object detection networks CSP-DarkNet-53, ResNet-50, and Tiny YOLO in detecting highly variable bottlenose dolphin (Tursiops truncatus) whistles using DeepAcoustics, a user-friendly deep learning tool. We compared the F1-score, average precision (AP), and mean AP performance of all network architectures with combinations of training samples from each acoustic environment. CSP-DarkNet-53 consistently outperformed Tiny YOLO and ResNet-50 across various test datasets, demonstrating robustness, but underperformed in select scenarios. Performance remained higher for aquarium data compared to open ocean data based on AP and mean AP values, indicating a greater ability of the networks to accurately detect whistles in these environments. However, networks trained on open ocean datasets showed only slightly improved APs on open ocean data, highlighting the challenge of achieving generalizability across divergent acoustic environments. This effort highlights the importance of network architecture selection, and the effects of different acoustic environments on deep learning methods for detecting complex underwater vocalizations.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Deep LearningVocalization, AnimalBottle-Nosed DolphinAcousticsAnimalsNeural Networks, ComputerSound SpectrographySignal Processing, Computer-Assisted

Autor:innen und Institutionen

  • Acoustic Interactions, LLC, Bellevue, Washington 98008, USA.
  • Ocean Science Analytics LLC, San Diego, California 92129, USA.
  • Integral Consulting Inc., Santa Cruz, California 95060, USA.
  • Stokes School of Marine and Environmental Sciences and Joan M. Sinnott Chair of Psychology, University of South Alabama, Mobile, Alabama 36688, USA.

ORCID-Profile: Elizabeth L Ferguson

Zitieren

APA

Sugarman, P. C., Ferguson, E. L., Alongi, G. C., Schallert, J. P., & Lyn, H. (2025). Effects of network selection and acoustic environment on bounding-box object detection of delphinid whistles using a deep learning tool. The Journal of the Acoustical Society of America. https://doi.org/10.1121/10.0036942

Vancouver

Sugarman PC, Ferguson EL, Alongi GC, Schallert JP, Lyn H. Effects of network selection and acoustic environment on bounding-box object detection of delphinid whistles using a deep learning tool. The Journal of the Acoustical Society of America. 2025. doi:10.1121/10.0036942

BibTeX
@article{Sugarman2025,
  author = {Sugarman, Peter C and Ferguson, Elizabeth L and Alongi, Gabriela C and Schallert, Jennifer Pettis and Lyn, Heidi},
  title = {Effects of network selection and acoustic environment on bounding-box object detection of delphinid whistles using a deep learning tool},
  journal = {The Journal of the Acoustical Society of America},
  year = {2025},
  doi = {10.1121/10.0036942}
}
Kostenlos starten →