In recent years, I’ve become a regular participant in the RSPB’s Big Garden Birdwatch, contributing my citizen science data to help track the UK’s garden birds. This year I took a different approach – relying not just on my eyes, but using BirdNET-Pi, which harnesses AI to identify bird calls from audio, creating a richer picture of the birds in my neighborhood.

BirdNET-Pi is an open-source project available on GitHub designed to run on a Raspberry Pi, a small and affordable computer. After setting up BirdNET-Pi, I can easily access a comprehensive dashboard via a web browser, where I can explore detailed information about the birds identified in my area from a microphone I snaked out my window.

Whilst there are countless guides on how to set-up BirdNET-Pi, I thought I’d share the details of my configuration.

The kit:

• Raspberry Pi 5 (4GB model is fine, however I opted for 16GB)
• Raspberry Pi 5 Active Cooler. To take the heat off crunching those chirps!
• Raspberry Pi 5 USB-C power supply
• Pibow Raspberry Pi 5 case. Because it looks pretty cool and nicely fits around the active cooler.
• SD card (for the OS image with ample capacity to store audio data – I purchased 128GB).
• USB microphone (I purchased a Movo M1 USB Lavalier microphone, a relatively inexpensive option).

The specific assembly instructions for the Pibow for Rasperry Pi 5 can be found at https://learn.pimoroni.com/article/building-your-pibow-5. The peripherals for Pi 5 can easily be swapped out for Raspberry Pi 4 (with the associated case, power supply and active cooler for 4).

I’m going to skip straight to BirdNET-pi, as there are guides aplenty on how to set-up Rasperry Pi. However the official guide was enough to get up and running: https://www.raspberrypi.com/documentation/computers/getting-started.html.

I installed BirdNET-Pi from a particular fork to ensure compatibility with the Raspberry Pi 5, which runs on the Debian Bookworm OS: https://github.com/Nachtzuster/BirdNET-Pi

The installation worked without a hitch, perhaps a moments delay before the data began to flow in and I was able to locate my installation on my local network under birdnetpi.local.

BirdNET-pi has support for Apprise – a notification library that can be configured under via the Settings menu. For example, I subscribed to MQTT (a type of message broker) to join a community network of BirdNET-Pi devices in my locality, enabling us to share bird detection data and collaborate on monitoring local bird populations.

So how about the Big Garden Bird Watch results? Herring Gulls, House Sparrows and Starlings amongst the most numerous detections were of no surprise. Undetected by BirdNET-pi were my more silent avian neighbours, a flock of Feral Pigeons and a pair Carrion Crows. However a couple of detections BirdNET-pi made irrespective of what I saw included the Eurasian Oystercatcher, Dunlin and Common Goldeneye. Oystercatchers and dunlins are regulars in my neighborhood, but I’ve never spotted a Goldeneye -though the audio recording sounded legit! And my most counted bird? The feral pigeon!

Not only was this is a novel way for doing the Big Garden Bird Watch this year, the motivation to get my own kit came after learning about the Dundee Bionet, a network of acoustic bird detectors spread across green spaces and community gardens in Dundee.

This is only just the beginning of my foray into bird audio data – stay tuned for what projects take flight next!