Twelve quick tips for applying deep learning to animal sounds
Burooj Ghani1, Anne Leonie Baier1, Vincent J Kalkman2
1Understanding Evolution Group, Naturalis Biodiversity Center, Leiden, The Netherlands.
Plos Computational Biology
|August 12, 2026
Summary
This guide offers practical steps for biologists to use machine learning in animal sound analysis. It provides a reproducible framework for developing and deploying bioacoustic models, making advanced tools accessible.
Area of Science:
- Bioacoustics
- Computational Ecology
- Machine Learning in Biology
Background:
- Deep learning advances automated analysis of animal sound for species identification, behavior, and ecological studies.
- Many biologists lack the computational expertise for implementing complex bioacoustic machine learning models.
Purpose of the Study:
- To provide practical, science-led guidelines for the full lifecycle of bioacoustic machine learning.
- To demystify software engineering for researchers applying machine learning to animal acoustics.
Main Methods:
- An iterative framework for building machine learning workflows.
- Leveraging transfer learning and open-source tools.
- Evaluating models based on real-world error costs and addressing domain shift.
Main Results:
- A low-barrier, reproducible pathway for researchers using machine learning on animal sound data.
- Guidelines cover problem definition, data annotation, model training, evaluation, deployment, and ethics.
Conclusions:
- Demystifies the software-engineering process for bioacoustic research.
- Enables wider adoption of machine learning in ecological and behavioral studies using sound.
