End to end workflows in Ecoacoustics: Verifying AI Output

Monitoring biodiversity at scale is critical for conservation, yet many of Australia s most threatened species are cryptic, rare, or inhabit hard-to-reach places. Traditional survey methods are often limited in reach and repeatability. Passive acoustic monitoring offers a practical, repeatable, and non-invasive way to detect presence, assess diversity, and monitor ecological change; all by listening to the soundscapes of Country. We at Open Ecoacoustics are continuously developing our Acoustic Workbench software, an open-source system that enables users to upload, annotate, and analyse large volumes of acoustic data. This software integrates AI models such as Google s Perch and Cornell s BirdNET to automate species identification. We work in partnership with NGOs, government agencies, and citizen scientists, supporting diverse ways of working. The Acoustic Workbench software has enabled long-term, landscape-scale monitoring across diverse ecosystems. In several case studies, acoustic data has revealed greater species diversity than traditional methods. Automated analysis has significantly reduced the time and expertise required to process recordings, supporting faster, evidence-based decision-making in conservation and land management. This presentation will cover improvements to our analysis workflows, particularly in relation to the verification of AI generated results. Open Ecoacoustics is a project supported by the Australian Research Data Commons (ARDC) through the Planet RDC program. This infrastructure is designed to support machine observation and data processing at scale. By integrating open platforms and AI, we are deepening ecological insight, detecting early signs of decline, and empowering communities to take informed action. ?