Acoustic AI Detection

Technology it uses

• Ultrasonic Microphones High-bandwidth (sub-1 Hz to >190 kHz) microphones for recording bat echolocation calls and other ultrasonic vocalisations. The primary capture channel for bat acoustic monitoring across Australian threatened-species work.
• Edge ML Inference Lightweight ML classifiers running on field hardware — classifying audio, video, or sensor data as it is captured. No internet required; decisions in seconds; alerting and curtailment-grade response possible at sites with no reliable connectivity. Used for bat call detection, starling sentinel work, and rare-event triggering across continuous audio.
• Bulk ML Inference High-throughput cloud GPU pipelines for processing archived recordings. This supports retrospective surveys, classifier comparison, and rare-event searches across years of SongMeter, Anabat, or Pettersson audio. We use custom classifiers for Australian threatened species alongside re-engineered baselines such as BirdNET and Google Perch.
• Field Platform The physical and electronic base that lets every monitoring system run unattended for months at a time. Custom Linux-based edge nodes with regulated solar power, satellite communications, hardened enclosures, integrated sensor I/O, and remote fleet management. Engineered for Pilbara heat and dust, NSW forests, cave humidity, and cyclone-rated installations.
• Dashboards & Review Tools Client dashboards provide access, replay, and report downloads. Analyst review tools support annotation, model comparison, and confidence review. Together they turn model results into information clients can use in approvals and operations.