A Raspberry Pi 5 listens to your backyard and draws the birds it hears, in real time, on an e-paper display. Builder arnegiacomo combined BirdNet-Go, an open-source AI birdsong classifier, with a USB microphone to identify species by audio alone. The identified birds appear as illustrations pulled from 19th-century naturalist guides, assembled into a live collage.
The hardware stack is straightforward: Pi 5 running BirdNet-Go, one USB mic, one large e-paper screen. The interesting engineering is in the pipeline connecting acoustic classification to image retrieval and layout. The project is live in Norway, and you can watch its current bird detections in real time at fugleramme.arnegiacomo.dev, where the species list reads as a credible cross-section of European fauna.
Read the full writeup for the GitHub source and to understand how BirdNet-Go slots into a project like this. The classifier is the part worth examining: it does the heavy lifting that makes the display more than a slideshow. This is also not the first audio-driven bird identification build Hackaday has covered, which means a comparison to prior approaches is waiting for anyone who digs into the linked archives.
[READ ORIGINAL →]