Guitar_Chord_Classifier

A machine learning model that classifies what guitar chord is being played and the quality of it: i.e if it's ringy, clear, or muted.

Data not included

Versions:

• Built in python3.

Libraries Required:

• NumPy
• PyDub
• FFmpeg (PyDub Dependency)

Usage:

• Clone this repository. git clone https://github.com/McCrearyD/Guitar_Chord_Classifier
• To ONLY process raw audio into spectrograms for training, run python3 -u convert_data.py inside the main directory.
• More instructions will follow as more features come out! Currently under development!

Adding Custom Data:

• Record your chord and note whether it's clear, ringy, or muted AND what exact chord you're playing.
  • For example, [G, ringy]
• Create a file in the root directory and name it raw_data if there isn't already one.
• Inside the raw_data directory, create a new directory for the chord name (again, if there isn't already one).
• Inside raw_data/${chord_name}, create yet another directory for the "quality" of the strum, in this case "ringy".
• You should have a directory set up similar to: raw_data/G/ringy in this example case.
• INSIDE the ringy sub-directory, append all files that are "ringy Gs".
• Repeat this process for any chords, qualities, etc. you wish to train on.

Note: All audio file types are accepted for raw data, as they will be converted into spectrogram images. Also all multi-channel audio files will be converted to 1.

FILE SIZES MUST BE A MINIMUM OF 2 SECONDS. ALL INPUT DATA WILL BE TRUNCATED FROM THE BEGINNING TO 2 SECONDS