An AI-enhanced bird feeder system using Raspberry Pi 4, Raspberry Pi Zero 2W, Google Coral TPU, and Raspberry Pi Pico W for bird detection and recording.
FeatherFeed is a smart bird feeder monitoring system that merges the beauty of nature with cutting-edge technology. The system uses Raspberry Pi 4, Raspberry Pi Zero 2W, Raspberry Pi Pico W, and Google Coral TPU to detect, record, and identify birds that visit the feeder. The system alerts users about bird visitations through a user-friendly web application.
For a comprehensive overview of how FeatherFeed works and its features, check out our project video on YouTube:
This video gives you a peek into the setup, functionality, and the technology behind this innovative bird feeder system.
Throughout the development of Feather Feed, we experimented with two different hardware configurations to optimise bird detection accuracy and system efficiency. Below are the descriptions and images of these configurations:
In this setup, the ultrasonic sensor and the camera were placed on either side of the bird feeder box. This configuration allowed for a wide detection range but presented challenges in aligning the sensor's detection with the camera's field of view.
After testing and evaluation, we found that placing both the ultrasonic sensor and the camera at the back of the box, pointing outwards, significantly improved the system's performance. This setup ensured that whenever a bird triggered the ultrasonic sensor, it was already in the camera's field of view, making for more efficient and accurate bird detection. This configuration will be the basis for Version 2 of the Feather Feed system.
To get your FeatherFeed system up and running:
- Assemble the Hardware: Follow the setup instructions for each device.
- Install the Software: Use the provided scripts to install necessary packages and dependencies.
- Configure the System: Set up the Mosquitto broker, Coral TPU, and sensors as per the documentation.
- Launch the Web App: Access the web application to start monitoring bird activity in real-time.
For a detailed guide, refer to the installation instructions for each device below.
- Raspberry Pico W: Equipped with an Ultrasonic Sensor, Temperature/Humidity Sensor, and LED light, it detects bird presence and publishes data on temperature, humidity, and visitor status.
- Raspberry Pi Zero 2W: Dedicated to video recording using the Arducam IMX519 16MP camera. It starts recording when a visitor is detected and stops when the visitor leaves.
- Raspberry Pi 4: Serves as the main processing unit. It uses the Google Coral TPU and TensorFlow compatibility, runs Mosquitto broker, and processes video files for bird identification and uploading.
- Real-Time Bird Presence Detection: Raspberry Pico W detects and publishes bird visitation data.
- Automated Video Recording: Raspberry Pi Zero 2W records video upon detecting visitors.
- AI-Powered Bird Identification: TensorFlow Lite and Google Coral TPU are used for accurate species identification.
- Instant Notifications: Updates users on bird visitations through an interactive web app.
- Raspberry Pi Pico W: Monitoring bird presence with sensors.
- Raspberry Pi Zero 2W: Video recording, running Raspbian 64bit Bullseye Lite.
- Raspberry Pi 4B 4GB: AI processing, MQTT broker, and communication hub.
- Google Coral TPU: Enhances TensorFlow’s capabilities for species identification.
- Operating System: Raspbian.
- Machine Learning: TensorFlow Lite.
- Communication: MQTT (Mosquitto) broker.
- Programming Language: Python, JavaScript.
- API: Built with Hapi framework.
- Data Storage: PostgreSQL using Supabase.
- Frontend: SvelteKit, Tailwind CSS, hosted on Vercel.
- Raspberry Pico W Setup: Instructions in
device/picodirectory. - Raspberry Pi Zero 2W Setup: Instructions in
device/zerodirectory. - Raspberry Pi 4 Setup: Instructions in
device/pidirectory.
Visit our website: featherfeed.ie