Bird populations play an important ecological role as they both mitigate pests and can also themselves be pests. Because of this, monitoring bird populations provides important insights into the overall health of an ecosystem with implications for both conservation and for agriculture.

Traditionally bird populations are monitored through catch and release tagging. (Fuller et al) This has the drawbacks of being resource intensive, is invasive to the birds, and introduces sampling bias as it is difficult to ensure uniform representation across species and ecosystems.

The goal of our project is to identify birds through their birdsong to facilitate population monitoring that can be done remotely, inexpensively, and can be easily automated. These techniques can also be expanded to discover new species by looking for songs that do not match any known species.

• EDA.ipynb

  This notebook contains metadata analyis of the audio dataset, audio analyis, dataset balancing and dataset preparation for export to Edge Impulse.

• Edge Impulse Project - https://studio.edgeimpulse.com/studio/419222

  This is where we did feature generation, model design exploration and packaging for Arduino

• edge-impulse-model.py

  This file is just the model code from Edge Impulse.

• ei-what-the-chirp -final-project-arduino-1.0.12.zip

  The Arduino sketch generated by Edge Impulse and then modified to interact with the LED display.

• WhatTheChirp.zip

  Our Arduio project that leverages the sketch as a library and then incorporate displaying inferences on a small screen as demonstrated in the video.

• WhatTheChirp_EEP_596_FinalVideo.mp4

  A video demonstrating post-deployment results followed by data processing, training, and compression of out ML model on Edge Impulse.

• WhatTheChrip _EEP_596_FinalProjectReport.pdf

  The final report on the project that summarized our problem, methodology, system design and results.