Hybrid Disagreement-Diversity Active Learning for Bioacoustic Sound Event Detection

TL;DR

This paper introduces MFFT, an active learning method combining disagreement and diversity analysis, significantly reducing annotation needs for bioacoustic sound event detection while maintaining high accuracy, especially for rare species.

Contribution

The paper proposes MFFT, a novel active learning approach that effectively reduces annotation effort in BioSED, particularly benefiting cold-start and rare species detection.

Findings

MFFT achieves 68% mAP cold-start, 71% warm-start, close to 75% fully-supervised.

Uses only 2.3% of annotations, reducing labeling cost.

Excels in detecting rare species and cold-start scenarios.

Abstract

Bioacoustic sound event detection (BioSED) is crucial for biodiversity conservation but faces practical challenges during model development and training: limited amounts of annotated data, sparse events, species diversity, and class imbalance. To address these challenges efficiently with a limited labeling budget, we apply the mismatch-first farthest-traversal (MFFT), an active learning method integrating committee voting disagreement and diversity analysis. We also refine an existing BioSED dataset specifically for evaluating active learning algorithms. Experimental results demonstrate that MFFT achieves a mAP of 68% when cold-starting and 71% when warm-starting (which is close to the fully-supervised mAP of 75%) while using only 2.3% of the annotations. Notably, MFFT excels in cold-start scenarios and with rare species, which are critical for monitoring endangered species, demonstrating its practical value.