Senior Comprehensive Project

Author: Grey Hutchinson

My project, which I nicknamed “Murmur”, was to create a research tool that would use neural networks, which can read and learn patterns to try and classify an audio file, to analyze any given PAM file and return a list of timestamps where the target species made a vocalization.

Note from Grey: Please contact me if you feel like you can’t grade this project without copious amounts of data. I get it, but also I can’t house it all on github. Not sure what to do here.

How to: Install this project and create your model:

Make sure you have python3.6 or greater installed on you computer
Create a virtual environment
1. python -m venv venv
2. Windows: ./venv/Source/activate
Install requirements.txt
1. pip install -r requirements.txt
Separate your files into those containing your target species and those that do not.
Place the files containing your target species into the “trues” folder; those without in the “falses” folder.
Run dataframe_maker.py
Run Murmur.py, if you’d like to create a new model. (Doesn’t work without data (duh, but, ya know))

How to: Run this project with passive acoustic monitoring files

Extract saved_model.zip to this directory, making sure the folder is called “saved_model”. This is my most recent model.
Save the PAM file to the root directory (e.g. pam_1.wav). If you’d like to just try my model, you can download this bit of relaxing beach sounds and use it as your PAM file: https://www.youtube.com/watch?v=QX4j_zHAlw8 It’s not too large.
Go into app.py and change line 131 to match your file name
Run app.py