BirdSet: A Large-Scale Dataset for Audio Classification in Avian Bioacoustics

2024/03/15 by Lukas Rauch, Raphael Schwinger, Rauch, Lukas +19 · 1 voice · 12 citations
Biochemistry, Genetics and Molecular Biology · Earth and Planetary Sciences · Environmental Science · #Animal Vocal Communication and Behavior #Marine animal studies overview #Underwater Acoustics Research #cs.AI #cs.SD #eess.AS

paper · pdf · doi:10.48550/arxiv.2403.10380

openalex publication_date 2024/03/15 · openalex created_date 2024/03/19 · openalex updated_date 2026/07/28

Abstract

Deep learning (DL) has greatly advanced audio classification, yet the field is limited by the scarcity of large-scale benchmark datasets that have propelled progress in other domains. While AudioSet is a pivotal step to bridge this gap as a universal-domain dataset, its restricted accessibility and limited range of evaluation use cases challenge its role as the sole resource. Therefore, we introduce BirdSet, a large-scale benchmark dataset for audio classification focusing on avian bioacoustics. BirdSet surpasses AudioSet with over 6,800 recording hours (\uparrow 17%) from nearly 10,000 classes (\uparrow 18×) for training and more than 400 hours (\uparrow 7×) across eight strongly labeled evaluation datasets. It serves as a versatile resource for use cases such as multi-label classification, covariate shift or self-supervised learning. We benchmark six well-known DL models in multi-label classification across three distinct training scenarios and outline further evaluation use cases in audio classification. We host our dataset on Hugging Face for easy accessibility and offer an extensive codebase to reproduce our results.

Cited by

• animal2vec and MeerKAT: A self-supervised transformer for rare-event raw audio input and a large-scale reference dataset for bioacoustics
• WhAM: Towards A Translative Model of Sperm Whale Vocalization
• Data-Efficient Self-Supervised Algorithms for Fine-Grained Birdsong Analysis
• Uncertainty Calibration of Multi-Label Bird Sound Classifiers
• Merlin L48 Spectrogram Dataset
• Unmute the Patch Tokens: Rethinking Probing in Multi-Label Audio Classification
• WoW-Bench: Evaluating Fine-Grained Acoustic Perception in Audio-Language Models via Marine Mammal Vocalizations
• What Matters for Bioacoustic Encoding
• Perch 2.0: The Bittern Lesson for Bioacoustics
• Foundation Models for Bioacoustics -- a Comparative Review
• Adversarial Training Improves Generalization Under Distribution Shifts in Bioacoustics
• Distilling Spectrograms into Tokens: Fast and Lightweight Bioacoustic Classification for BirdCLEF+ 2025

Discussions

• Just learned about another dataset arxiv.org/abs/2403.10380 We're going to combine BirdSet with iNatSounds to make a new, even bigger, more strongly labeled dataset, that we call iBirdSuperSet, and [bsky, 6 points, 2 comments]

Related

• Can Masked Autoencoders Also Listen to Birds?
• Auto deep learning for bioacoustic signals
• AudioProtoPNet: An interpretable deep learning model for bird sound classification
• Explainability of CNN Based Classification Models for Acoustic Signal
• What Matters for Bioacoustic Encoding