Stratified Active Learning: Release v1.1.0

Zenodo (CERN European Organization for Nuclear Research)

Abstract

This repository contains the implementation of stratified active learning methods for the paper Stratified Active Learning for Spatiotemporal Generalisation in Bioacoustic Monitoring. Key Features: Multiple Stratification Methods: Spatial (location-based) and temporal (time-based). Uncertainty Quantification: Binary entropy, ratio max, and cluster-based uncertainty measures Active Learning Pipeline: Complete implementation with batch sampling and accumulation strategies Evaluation Framework: Per-strata performance evaluation using mAP, cmAP, and F1 scores with the Pielou Index and Coefficient of Variation to measure evenness across strata. Full Changelog: Added compare_methods.ipynb to generate combined strata comparison figures. Clarified documentation and included links to datasets

Authors 3

1. Ben McEwen

   Tilburg University

   Affiliation as printed

   Tilburg University

2. Corentin Bernard

   Laboratoire d’Informatique et Systèmes

   Affiliation as printed

   Laboratoire d'Informatique et Systèmes

3. Dan Stowell Aachen Naturalis Biodiversity Center

   Tilburg University · Naturalis Biodiversity Center

   Affiliation as printed

   Tilburg University

   Naturalis Biodiversity Centre