Acoustic monitoring from detection to decision

MANDY evolves passive acoustic bird monitoring from a collection of automated detections to a structured, reproducible system for biodiversity assessment. BirdNET or BirdWeather exports are converted into species-specific activity heatmaps, confidence diagnostics, classification-specific overviews and review-oriented reports.

The project addresses a central bottleneck of acoustic monitoring: to assess large streams of machine-generated evidence in a transparent, fast and scalable way without losing the link to technical validation.

A pipeline for scalable acoustic biodiversity intelligence

At LIB, MANDY is being built as a modular analysis environment that links automated data import, visualisation, rule-based pattern scoring and quality control. The pipeline calculates Pattern Plausibility Scores (PPS), Artifact Suspicion Scores (ASS), preliminary classification labels, review tables and class-specific heatmap collections.

A key development goal is to move beyond static summaries. Sliding-window tests are being used to scan activity heatmaps systematically, detect temporally concentrated patterns and prepare the ground for automated quality assessment of continuous acoustic datasets.