knowSound( )

knowSound( )

Oct 7, 2019
IN PROGRESS

I'm planning to train a sound classification model for ml5 among a group of CV models. I've found a relatively small (around 1000-2000 pieces of sound effects files) yet well labeled dataset of sound effects: Adobe Audition Sound Effects. The dataset, or the sound effects library, is high quality while open-sourced and royalty-free. All the files were well named thus perfectly labeled for the training, here're several examples:

>>> Animal
Animal Dog Bark 08.wav
>>> Crash
Crash Car Crushing 01.wav
>>> Horror
Horror Wood Crackle 02.wav
>>> Sports
Sports Boxing Gloves Hit 03.wav

For every category, the most files were named by “Category” + “Object” + “event”, which is perfect for multi-layer training. I'll start looking for more references and writing the code this week.

References

  1. Sound Classification using Deep Learning
  2. Using Deep Learning for Sound Classification: an In-depth Analysis
  3. Enhanced Environmental Sound Classification with a CNN
  4. A Quick Introduction to the “Pandas” Python Library