Generalization in birdsong classification: impact of transfer learning methods and dataset characteristics

arXiv (Cornell University)

Abstract

Animal sounds can be recognised automatically by machine learning, and this has an important role to play in biodiversity monitoring. Yet despite increasingly impressive capabilities, bioacoustic species classifiers still exhibit imbalanced performance across species and habitats, especially in complex soundscapes. In this study, we explore the effectiveness of transfer learning in large-scale bird sound classification across various conditions, including single- and multi-label scenarios, and across different model architectures such as CNNs and Transformers. Our experiments demonstrate that both fine-tuning and knowledge distillation yield strong performance, with cross-distillation proving particularly effective in improving in-domain performance on Xeno-canto data. However, when generalizing to soundscapes, shallow fine-tuning exhibits superior performance compared to knowledge distillation, highlighting its robustness and constrained nature. Our study further investigates how to use multi-species labels, in cases where these are present but incomplete. We advocate for more comprehensive labeling practices within the animal sound community, including annotating background species and providing temporal details, to enhance the training of robust bird sound classifiers. These findings provide insights into the optimal reuse of pretrained models for advancing automatic bioacoustic recognition.

Authors 5

1. Burooj Ghani Aachen Naturalis Biodiversity Center

   Naturalis Biodiversity Center

   Affiliation as printed

   Naturalis Biodiversity Center , Leiden , The Netherlands

2. Vincent J. Kalkman Aachen Naturalis Biodiversity Center

   Naturalis Biodiversity Center

   Affiliation as printed

   Naturalis Biodiversity Center , Leiden , The Netherlands

3. Willem-Pier Vellinga Aachen Burooj Ghani Naturalis Biodiversity Center

   Naturalis Biodiversity Center

   Affiliation as printed

   Burooj Ghani , Naturalis Biodiversity Center , Leiden , Netherlands

   Xeno-Canto Foundation , The Netherlands

4. Lisa Gill
   Affiliation as printed

   Landesbund fur Vogel-und Naturschutz , Hilpoltstein , Germany

5. Dan Stowell Aachen Naturalis Biodiversity Center

   Naturalis Biodiversity Center · Tilburg University

   Affiliation as printed

   Department of Cognitive Science and Artificial Intelligence , Tilburg University , Tilburg , The Netherlands

   Naturalis Biodiversity Center , Leiden , The Netherlands