MetaPerch: Learning from metadata for bioacoustics foundation models

2026-07-15

AI Summary

Existing bioacoustic foundation models struggle to generalize across species distributions and acoustic domain shifts due to their neglect of rich metadata present in citizen science recordings. This work presents the first systematic investigation into the impact of nine metadata categories—such as geographic location and recording time—on representation learning. To address this gap, we propose MetaPerch, a multitask learning framework that leverages metadata as auxiliary supervision signals to jointly model acoustic features and environmental context, thereby enhancing robustness and generalization in species identification. Extensive experiments across 17 bioacoustic datasets demonstrate that our approach significantly improves cross-domain recognition accuracy, validating the effectiveness and broad applicability of metadata-augmented learning for bioacoustic foundation models.

Abstract

Bioacoustic foundation models rely on large-scale citizen science platforms like Xeno-Canto for geographically and ecologically diverse data. Recent work has shown that supervision alone can produce SotA species detection models when trained on this large-scale data -- however, there remains unutilized potential in the form of recording metadata readily available within these community-driven data hubs. In this work, we explore the use of metadata -- such as location and time -- as auxiliary supervision signals, allowing the model to leverage species-metadata correlations in its learned representation. Auxiliary metadata losses provide additional information beyond vocalizations alone that can encourage a richer, more robust representation that generalizes better to species distribution and acoustic domain shifts -- important challenges for deployment in real-world passive acoustic monitoring (PAM) settings. We introduce MetaPerch, a new foundation model that achieves strong species identification performance across multiple challenging domains and present an extensive empirical study of the effects of 9 diverse metadata sources on 17 bioacoustic datasets.

Problem

• bioacoustics
• foundation models
• metadata
• species identification
• domain shift

Innovation

• metadata supervision
• bioacoustic foundation models
• auxiliary learning
• domain generalization
• passive acoustic monitoring