Music Genre Classification Project

Overview

Project exploring machine learning approaches for automated music genre classification using the FMA dataset and metadata.

Repository Structure

• /code/: Python implementations for genre classification
  • BasicModels/: Baseline classification model implementations using sklearn libraries
  • Processing/: Data preparation scripts
  • CombinedFeatures/: Models using multiple feature types for experiments
  • SingleFeatures/: Single feature type experiments
  • Graphs/: Visualization generation
• /fma_dataset/: Audio data and metadata
• /Preliminary Results/: Performance graphs and genre mappings

Setup

1. Download FMA dataset and verify checksums
2. Run initialPreprocessing.py from Processing/
3. Run any file in CombinedFeatures or SingleFeatures to generate results (install dependencies per file)

Results

• Full experiment results: Google Sheets
• Technical paper: PreliminaryResults/MusicGenreClassification.pdf

Contributors

• Catherine Baker
• Thomas Davidson