Perch 2.0: The Bittern Lesson for Bioacoustics

About

Perch is a performant pre-trained model for bioacoustics. It was trained in supervised fashion, providing both off-the-shelf classification scores for thousands of vocalizing species as well as strong embeddings for transfer learning. In this new release, Perch 2.0, we expand from training exclusively on avian species to a large multi-taxa dataset. The model is trained with self-distillation using a prototype-learning classifier as well as a new source-prediction training criterion. Perch 2.0 obtains state-of-the-art performance on the BirdSet and BEANS benchmarks. It also outperforms specialized marine models on marine transfer learning tasks, despite having almost no marine training data. We present hypotheses as to why fine-grained species classification is a particularly robust pre-training task for bioacoustics.

Bart van Merriënboer, Vincent Dumoulin, Jenny Hamer, Lauren Harrell, Andrea Burns, Tom Denton · 2025

Related benchmarks

Task	Dataset	Result
Bioacoustic Classification	CBI	Accuracy79.2	10
Bioacoustic Classification	BirdSet	AUROC0.908	8
Active Labeling	Watkins	Fraction of Rare Samples Surfaced62	6
Active Labeling	CBI	Fraction of Rare Samples Surfaced81	6
Active Labeling	EFB	Fraction of Rare Samples Surfaced69	6
Active Labeling	HumBugDB	Fraction of Rare Samples Surfaced49	6
Active Labeling	Dogs	Fraction of Rare Samples Surfaced91	6
Active Labeling	Kruger	Fraction of Rare Samples Surfaced78	6
Active Labeling	Camdeboo	Fraction Rare Samples Surfaced73	6
Active Labeling	Kgalagadi	Rare Sample Fraction76	6