README.md

MIDI Datasets: Rock, Pop, and Game Themes

Dataset Overview

This repository contains three curated MIDI datasets focusing on different music genres and themes. Each dataset was meticulously prepared from publicly available compositions on the internet, cleaned, and filtered to ensure high-quality training data for machine learning models. These datasets are suitable for tasks such as music generation, genre classification, and other creative applications.

The datasets included in this repository are:

1. Rock MIDI Dataset: Contains approximately 2,500 compositions.
2. Pop MIDI Dataset: Contains approximately 1,000 compositions.
3. Game Themes MIDI Dataset: Contains approximately 200 compositions.

Dataset Preparation

The datasets were prepared following the same workflow:

1. Collection: MIDI files were gathered from publicly available sources on the internet.
2. Cleaning: Duplicate files were removed, formatting issues were corrected, and compositions were checked for completeness.
3. Filtering: Only high-quality compositions relevant to the specific genre or theme were retained.

Each dataset offers a unique focus, providing a rich source of material for training music-related machine learning models.

Dataset Details

1. Rock MIDI Dataset

• Size: Approximately 2,500 compositions.
• Description: Focuses on the rock genre, including a diverse range of rock substyles. Suitable for applications aiming to generate or analyze rock music.

2. Pop MIDI Dataset

• Size: Approximately 1,000 compositions.
• Description: Focuses on the pop genre, providing a collection of melodic and harmonic compositions typical of pop music.

3. Game Themes MIDI Dataset

• Size: Approximately 200 compositions.
• Description: A niche dataset focusing on iconic and original MIDI compositions related to game themes, suitable for gaming-related music applications.

Usage

These datasets are ideal for:

• Training models for music generation.
• Genre classification and analysis.
• Style transfer and creative applications.

License

Please ensure compliance with the licensing terms of the sources from which the compositions were gathered. These datasets are provided "as is" without any guarantees or warranties.