Music Classification Mini-Project

Given Midi music files, predict the artist.

Project Objective

The primary aim of this project is to build a machine learning classifier capable of identifying whether a given MIDI file was composed by one of four classical composers: Bach, Beethoven, Schubert, or Brahms. The classifier should be able to process audio data in intervals of 15, 30, or 60 seconds, with a default setting of 30 seconds.

Data

The dataset consists of several hundred annotated MIDI files, corresponding to compositions from the four composers. The data is provided in data/raw/training/ and is originally sourced from Musicnet.

Data Format

• MIDI files
• Composition name precedes the first underscore in the file name
• Composition artist (target) is given in as the parent folder name

Solution, Methods, and Results

• For an overview of the problem, solution, methods, and results, please see the project presentation.
• For specific details on the implementation, please see the Jupyter Notebooks for data creation and preprocessing and model training and evaluation.
  • Alternatively, these notebooks have been as HTML files to view locally.

Quickstart Guide

1. Clone the Repository: Run git clone <repository_url> to clone the project.
2. Run Setup Script: Navigate to the project directory and execute ./setup.sh. This creates a Python virtual environment named my_venv with proper dependencies.
3. Activate Environment and Run Jupyter Notebook:
   • Option 1: Execute source my_venv/bin/activate followed by jupyter notebook.
   • Option 2: Use an IDE such as VS Code to activate the environment and run the Jupyter Notebook.