If you've ever wanted to know exactly what bird is near you by only the sound, then I have the solution for you: a fully-fledged bird AI system that runs completely on a Raspberry Pi single board computer. Hey gang, Tim here at Core Electronics and today we're setting up a Raspberry Pi to run BirdNet Lite. 

The system we're building here can run 24/7 non-stop. It records, tracks, and displays all recognized bird calls. You can access the data and audio live streams through any locally connected device or computer. It even saves the best recordings so you can listen to them later.

BirdNet Lite is the TensorFlow Lite version of the original BirdNet system. These are pre-trained machine-learned softwares that can recognize more than 6,000 species from only their sounds. On the table before me is everything you need to set up your BirdNet Pi system. Naturally, you're going to need a Raspberry Pi single board computer. We are using a Raspberry Pi 4 Model B 2 Gigabyte here. You're also going to need a microSD and some heat sinks for your Raspberry Pi, a USB microphone so you can capture the ambient sounds, a Raspberry Pi power supply, a micro HDMI to HDMI cord, a mouse, and a keyboard. I also 3D printed a case that you can see at the end of this video, but any passive or actively cooling cases would be a fantastic addition to this setup.

First step is to set up the Raspberry Pi as a desktop computer. Insert your microSD card into a computer (you may need an adapter to do so) and open up the Raspberry Pi Imager. Check the description for a download link for Raspberry Pi Imager. When navigating through the Imager, you must click the Raspberry Pi OS (other) to find and select the Raspberry Pi OS 64-bit version. Then select your connected microSD card and start the flashing process by clicking right. Next, attach your USB microphone to the USB.

Connect a 3.0 port to the Raspberry Pi to listen to ambient noises. Attach a heatsink to the top of the Raspberry Pi to keep it cool. Connect a flashed SD card, mouse, and keyboard to the system. Connect an internet-connected ethernet cord to the system.

Power up the system and you will be welcomed by the Raspberry Pi desktop. To install Birdnet Light onto the Raspberry Pi, open a new terminal window by pressing the black button on the top left of the screen. Copy and paste the curl command from the full written up article link to it in the description. This command will install all the required packages for the bird call identifying and recording system. If ever prompted by the terminal with the question "Do you want to continue?", press "y" and then enter key to continue the process. This command will take around 10 minutes. Once complete, the Raspberry Pi OS will reboot itself.

Pull up the Birdnet Local website that the Raspberry Pi automatically creates upon boot up. Copy the address for the user interface from the full article and paste it into the URL of any locally connected internet browser (Chrome or Firefox are excellent choices). Test the system by playing some magpie calls to the Raspberry Pi. This is a pre-trained system and the Raspberry Pi has never heard this recording before. It will figure out what bird that was all based on new information.

Refresh of our Bird Net Pie.local page and there is our Magpie with a confidence of 75. Let's see if we can do the same with the Kookaburra. A quick refresh of the page and there is our Laughing Kookaburra. Now here is a microSD that I've prepared before. I have had a Bird Net Pi system run for a couple of days on this particular Micro SD. Exploring the wealth of data and pages is going to be a lot more interesting with more identified species.

I'm going to quickly swap out the current microSD with this one and turn on the system. With that done, opening up the Bird Net webpage you can see many more birds have been identified. Diving into the data you can see the time of day each bird call happened, the identified bird call recorded, a spectrograph of all the sounds, images of the identified species, tables, charts - it is really all there.

Certain sections may require a username and password. By default the username is BirdNet and just leave the password empty. You can change these to if so desired. Worth mentioning if you only want a single species or if you want to exclude species from being identified you can do so in the tool section right here. This way if you don't want to track certain species or only track a single species your system is going to be easy to customize and good to go quickly.

If you want to participate in some citizen science then this is the project for you. As you watch this around the world there are many Raspberry Pies running Bird Net Pie right now. Some have even been specially set up so you can tune into them from anywhere in the world. With enough of these devices set up in a city or country the potential to figure out the flight paths and migrations of bird species in real time would be right at our fingertips. Bird Weather is another great example of citizen science at the grassroots stage - just check out.

That budding around the world coverage all the nodes have bird net systems running and the data you can see currently is for the last 24 hours. That's a lot of identifications for a single day! This is all possible thanks to everyone who's worked on anything birdnet related. Importantly for this project, a huge thanks goes to Patrick McGuire and everyone in the community who's put the hard yards in to make installing and running this system so effortless on a Raspberry Pi single board computer.

Now this software is all open source, so if you feel like some features are missing there's nothing stopping you from adding them. Go check out the GitHub for Bird Net Pi and the Bird Net Pi website links in the description. I also 3D printed a case which disguises the setup in the form of a garden rooster. The Raspberry Pi and a more omni-directional USB microphone tucks into it very nice and safe inside. This also provides some much-needed semi-waterproofing for outdoor placement. There are also some holes drilled into the case so the audio waves can still get to our AI system through the connected USB microphone.

That's it for today. If you have any questions write me a message below. We are full-time makers and we are here to help. So with my Bird Net Pi system fully functional and you with all the information to do the same, until next time stay cozy!