BirdSearch

Embedded device (breadboard, real product would be smaller)
Screenshot of the frontend dashboard

Inspiration

We were inspired by the need to create some application/product that uses embedded systems and machine learning with the goal to help the public in some way.

Over the past century, the world has undergone a trend of rapid urbanization. Large swaths of land have been transformed from untouched nature to dense human settlements.

It is important that these changes are undergone in an ecologically responsible manner. This requires accountability, for which affordable automated tools are most convenient.

What it does

Our project provides city planners with an affordable and scalable tool for monitoring the diversity of bird species in urban spaces. Small, low-cost, low-power edge devices collect ambient audio data and stream it to a cloud server, which uses a state-of-the-art AI model to classify bird calls by species. An online dashboard visualizes the distribution of species detected on a map, helping policymakers understand how their decisions impact biodiversity in the areas they influence.

How we built it

We use microcontrollers with a microphone to send audio data to a VULTR VPS server. The server then recognizes the audio data to determine the bird species using an AI classifier, and logs results to a MongoDB database. This information can be used to create an ecological map of the city to understand the variety of bird species and if any interventions are needed to change it to be more diverse and get rid of invasive species.

Challenges we an into

Initially we had planned to create some sort of system that can help map out pests like rats, but due to insufficient data our project scope had to change, but the core idea remained the same.