Inspiration
I think the research done by groups like Elephants Voices to try and translate elephants language was interesting, so I wanted to work on this project to try and eliminate some background noise from the audio.
What it does
The user uploads audio of an elephant call with some overlapping vehicle, plane, or generator in the background and receives a cleaned audio file along with a spectrogram comparison of the two audio files.
How we built it
I built the project using Flask, html, css, javascript, bootstrap, and other python libraries that can be found in my requirements.txt file. I used a simple boilerplate flask template to start with, and I used OpenAi to help me calculate metrics to measure the voice cleaning by and to work on my harmonic cleaner file that transforms an inputted audio file to a cleaned output file that aims to isolate the elephant voice.
Challenges we ran into
I wanted to use Auth0 at first to make people have accounts but I was given a tenant with a bad domain name that I couldn't change, and the free credits only allow one free build. I struggled in attempts to create a model that could guess what elephant call type it was just based off the input audio file because most of the files were just rumbles. Also trying to set up the harmonic cleaner file was hard too even with OpenAi's assistance.
Accomplishments that we're proud of
I completed my first Hackathon.
What we learned
Come with a group to my next hackathon, but don't stress too much about not being prepared.
What's next for ClearTusk
The next logical step is to move storage from static files in the project directory to storage like in the cloud.
Log in or sign up for Devpost to join the conversation.