Teddy Warner's Avian Visitors project identifies backyard bird species using a Raspberry Pi, a USB microphone, and BirdNET, Cornell University's deep learning classifier trained on over 11,000 species worldwide. The entire hardware requirement is minimal. The identification engine does the heavy lifting.
Detection data feeds a local web interface and, optionally, a color E-Ink display showing a 24-hour visitor collage where bird image size scales with detection frequency. The system also pushes detections to Home Assistant or over MQTT, enabling automation triggers keyed to specific species. That last detail is what separates this from a novelty.
The GitHub repository contains the full build. The original article is worth reading for Warner's implementation choices, specifically how regional reference image sets are pulled and composed into the collage, and how the E-Ink display pipeline is structured. The embedded build video makes the scope of the project concrete in under ten minutes.
[READ ORIGINAL →]