BirdNET-Pi is a real-time acoustic bird classification system designed for Raspberry Pi. It uses a USB sound card to pick up bird sounds, and classifies them locally using a pre-trained machine learning model. But the best bit is how the audio and visual data is shared around the world.

Developer Patrick McGuire knows of at least fourteen countries with live BirdNET-Pi installations, from Australia to Austria and from South Africa to Scotland.

What you need to build your own bird classifier

• Raspberry Pi 4 Model B or Raspberry Pi 3 Model B+
  
• SD card with the 64-bit version of Raspberry Pi OS Bullseye installed
  
• USB microphone or sound card

A comprehensive installation guide is available.

BirdNet-Pi installations around the world

Once you have everything set up, you need to join BirdWeather. It’s a “living library of bird vocalisations”, and lets you travel a map of the world looking and listening in to whatever species are in the area.

Simply scroll around the world map to see where BirdNET-Pi installations are live. Some setups only share acoustic data, which you can click to listen to, but lots also share live video.

We took the BirdWeather express to Laikipia in Kenya and saw another wingless species which looked a lot like a monkey. However, we do like monkeys, and we could also hear scores of birds calling out over the video’s audio.

As you can see from the pinned locations on the BirdWeather map below, most BirdNET-Pi installations are in North America and Europe. But we spotted one down in New Zealand, a few out in Japan, one in China, and handful in Africa and South America. It’s a truly live, global birdwatching experience.