Melodies Dataset
This directory contains the training, validation, and test datasets for melody generation.
Directory Structure
train/
(80% of data): Training datasetval/
(10% of data): Validation datasettest/
(10% of data): Test dataset
File Formats
- MIDI files (.mid)
- Each file follows naming convention:
[genre]_[mood]_[key]_[tempo].mid
Example:pop_happy_cmajor_120.mid
Genre Categories
- pop/
- jazz/
- classical/
Metadata Format
Each MIDI file should have corresponding metadata:
- Genre
- Key signature
- Tempo (BPM)
- Mood/emotion
- Time signature
Data Sources
[To be populated with actual sources]
Usage
This dataset is used for:
- Training melody generation models
- Model evaluation
- Style transfer experiments
File Count
- Train: [number] files
- Val: [number] files
- Test: [number] files
Notes
- All MIDI files are quantized to 16th notes
- Tempo range: 60-180 BPM
- All melodies are monophonic
Directory Structure
datasets/melodies/ βββ README.md βββ train/ β βββ pop/ β βββ jazz/ β βββ classical/ βββ val/ β βββ pop/ β βββ jazz/ β βββ classical/ βββ test/ βββ pop/ βββ jazz/ βββ classical/