Discerning levels of water through sound. Currently, the first notebook regarding collecting the audio data is released. From 5 samples of voice recordings of water filling up in different container, having two classes of outcomes for the model to predict - filling or overflowing. As the notebook was done in Google Colab, the code needs to be refactored for use in the PC.
The audios are split into 2 sec clips and then are converted to mellog images. Made use of librosa, pydub, sf, and other common ML libraries.