Intro to BirdNET-Pi: Eavesdropping on my feathered friends (hannahilea.com)

BirdNET-Pi is a low-cost, community-maintained application that runs Cornell Lab of Ornithology’s BirdNET classifier on a Raspberry Pi to do continuous, passive acoustic monitoring of birds. The author describes building multiple “giftable” units (Pi Zero 2 W + inexpensive mic) that record 15-second audio clips, convert them to spectrograms, and feed them to the BirdNET model to detect bird species; detections (and corresponding spectrograms/recordings) are stored locally and presented through a small web UI that visualizes daily summaries and recent events. The write-up emphasizes accessibility — clear step-by-step setup, tweaks to make Pi Zero installations stable over long runs, and community support via the Nachtzuster fork — making bird monitoring easy for hobbyists and citizen scientists. Technically, BirdNET-Pi pipelines continuous audio -> spectrogram -> model probabilities; it filters out clips with human-voice probability above a configurable threshold to protect privacy, and can optionally publish detections externally (disabled by default). BirdNET itself is CC BY-NC-SA and distinct from Merlin’s models. The project’s significance lies in democratizing ecoacoustics: low-cost, reproducible deployments enable neighborhood-scale biodiversity tracking, long-term behavioral observation, and community science, but users must weigh license limits and always-on microphone privacy trade-offs.