Instructions_GTZAN.md

How to Download the GTZAN Dataset

This project requires the GTZAN Genre Collection dataset. You can download it automatically using the provided Python script (recommended for TSOAI courses) or manually.

Option 1: Automatic Download (Recommended)

This method uses the dataset_downloader.py script to fetch the data directly from Hugging Face and organize the folders according to the project requirements. No API keys are required.

1. Install Requirements

This script uses standard Python libraries, so usually no extra installation is needed. If requests is missing:

pip install requests

2. Run the Script

Run the script to download, extract, and clean up the compressed files automatically:

python dataset_downloader.py

What this script does

1. Downloads the dataset (genres.tar.gz) directly from the Marsyas/GTZAN Hugging Face repository.
2. Creates Directory: It creates a ./GTZAN_dataset folder in your project root.
3. Extracts: Unzips the contents directly into that folder.
4. Cleanup: Automatically removes the .tar.gz file after extraction to save space.

---

Option 2: Manual Download

If the script fails or you prefer to manage files yourself:

1. Download

1. Visit the Hugging Face file repository: https://huggingface.co/datasets/marsyas/gtzan/tree/main/data
2. Download the file named genres.tar.gz.

2. Extract and Organize

1. Create a folder named GTZAN_dataset in your project root.
2. Extract the contents of genres.tar.gz into that folder.

3. Verify Directory Structure

Your project directory should look exactly like this for the code to run correctly:

DLforAudio/
├── GTZAN_dataset/                 
│   └── genres/         
│       ├── blues/            <-- For every genre, there is a folder.
│       │   └── blues.00000.wav 
│       └── ...
└── dataset_downloader.py

4. Path Configuration

Ensure your preprocessing scripts point to the correct folder location:

# preprocess.py (or your config file)
DATASET_PATH = "GTZAN_dataset/genres/"