At the suggestion of a couple of friends, I finally scraped together enough parts from the junk drawer to put together a BirdNET-PI system. This is a Tensorflow-based machine learning system that listens to the audio stream from an outdoor microphone and can identify over 6000 species. A station can be set up inexpensively; all that is needed for a minimal system is a Raspberry Pi board and an outdoor microphone.

This new station’s results can be seen on Birdweather:
https://app.birdweather.com/stations/7366

The results are pretty impressive so far, although I’ve seen a few false detections- the local squirrels are being detected as Belted Kingfishers.

The system is configured to ignore detections of Northern Saw-whet Owl, Boreal Owl, and Western Screech-owl for now since the owl banding operation uses audio lures of those species- no need to pollute the BirdNET results with our lure playback.