Ornimetrics — A Raspberry Pi Smart Bird Feeder Powered by AI Vision
The Project
OrniMetrics is an AI language and vision powered smart bird feeder that identifies visiting bird species in real time and autonomously responds. When a bird lands, the system dispatches the right food portion for that species and allows specific species to be blocked remotely through a companion app. It is a fully closedloop device: detection triggers a physical response every time. The inspiration was straightforward. I kept seeing birds at my feeder and had no idea what they were. I wanted to build something that didn't just passively log data but actually acted on the data it had, so it is not collecting data passively but actively using it. The biggest challenge was training the YOLOv11 model to perform reliably in real outdoor conditions with variable lighting, partial occlusions, and birds that don't stay still. The dataset grew to over 45,000 images before accuracy reached a usable level. On the software side, architecting the Flutter companion app with 13 service singletons across 6 feature tabs and maintaining Firebase sync real-time and race condition free took significant iteration of the app. The final layer was integrating AI to generate personalized biodiversity reports from detection history, turning raw species logs into useful natural language based responses. The full stack spans across a Raspberry Pi 5 8gb running local inference, Firebase for real time sync and authentication, a crossplatform Flutter app on iOS and Android, and an AI reporting layer.
