Art Or Fart?
Inspiration
We wanted to make positive change to the world, but thought that would be too hard, so we decided that it would be much easier to make something weird that we could be sarcastic about in the case we didn't like it.
What it does
Takes an audio file, performs a Fourier transformation, turns it into a spectrogram and analyses its frequencies to classify a sound as music or farting. Users can also upload their own sounds using the web interface.
How we built it
We started by finding 500 farts and converting them to spectrograms which we stored in txt files. We next started on the web interface by randomly generating a fart and song clip to find the confidence values of, as well as building the legendary Fartometer. We continued by adding the functionality to record audio on the webpage. Next, we used dimensionality reduction to find our reduced feature vectors, as well as labelling training data for later use. Next, we implemented K-Nearest Neighbour classification to complete our classifier; finishing up on Sunday by clearing up several aspects of the code.
Challenges we ran into
Challenge 1: Finding 500 clips of farting leads you to weird places. Challenge 2: Feature Selection has been extremely difficult Challenge 3: Attempting to record audio over the web interface was very difficult Challenge 4: It takes an eternity to load
Accomplishments that we're proud of
IT WORKS!!!!!!!!!!!!!!!!!!!! (sort of..)
- Learned to use Streamlit
- Pretty good time management and teamwork throughout
What we learned
- Over-engineering is a very real issue and nearly killed us
- How to use streamlit to create a GUI
- How to perform Fourier transforms in Python
What's next for Art or Fart - An Exercise in Overcommitment to the Bit
Hopefully nothing
Built With
- python
- streamlit
Log in or sign up for Devpost to join the conversation.