Can Masked Autoencoders Also Listen to Birds?

2025-04-17

AI Summary

General-purpose Masked Autoencoders (MAEs) struggle to capture domain-specific acoustic features critical for fine-grained avian vocalization classification in bioacoustic monitoring. Method: We propose Bird-MAE, the first MAE architecture specifically designed for avian acoustics. Leveraging the large-scale BirdSet dataset, we introduce an acoustic-aware MAE framework with optimized masking strategies, domain-adaptive pretraining, and novel frozen-representation utilization. We further devise prototypical probing—a parameter-efficient, highly discriminative transfer method for frozen representations. Results: Bird-MAE achieves state-of-the-art performance across all BirdSet downstream tasks. In multi-label classification, it significantly outperforms generic Audio-MAE baselines in mean Average Precision (mAP). Prototypical probing yields up to 37% higher mAP than linear probing and approaches full-parameter fine-tuning performance, with an average gap of only ≈3%.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Representation Learning for Vision
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Web Mining and Content Analysis: Large pretrained models with web data
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Masked Autoencoders (MAEs) pretrained on AudioSet fail to capture the fine-grained acoustic characteristics of specialized domains such as bioacoustic monitoring. Bird sound classification is critical for assessing environmental health, yet general-purpose models inadequately address its unique acoustic challenges. To address this, we introduce Bird-MAE, a domain-specialized MAE pretrained on the large-scale BirdSet dataset. We explore adjustments to pretraining, fine-tuning and utilizing frozen representations. Bird-MAE achieves state-of-the-art results across all BirdSet downstream tasks, substantially improving multi-label classification performance compared to the general-purpose Audio-MAE baseline. Additionally, we propose prototypical probing, a parameter-efficient method for leveraging MAEs' frozen representations. Bird-MAE's prototypical probes outperform linear probing by up to 37% in MAP and narrow the gap to fine-tuning to approximately 3% on average on BirdSet.

Problem

• Improving bird sound classification for environmental monitoring
• Adapting Masked Autoencoders for bioacoustic domain challenges
• Enhancing multi-label classification performance with Bird-MAE

Innovation

• Domain-specialized Bird-MAE for bioacoustic monitoring
• Pretrained on large-scale BirdSet dataset
• Prototypical probing improves frozen representation usage