🎼Datasets and projects for music generation and analysis tasks

MAESTRO - (MIDI and Audio Edited for Synchronous Tracks and Organization) contains over 200 hours of annotated recordings of international piano competitions over the past ten years.

NSynth - the dataset consists of 305,979 musical notes and includes recordings of 1006 different instruments, such as flute, guitar, piano and organ. The dataset is annotated by instrument type (acoustic, electronic or synthetic) and other sound parameters.

Lakh MIDI v0.1 - There are 176,581 MIDI files in the dataset, of which 45,129 are associated with samples from the Million Song Dataset. This dataset is designed to simplify the search for music information based on text and audio content on a large scale.

Music21 - contains musical performances from 21 categories and is aimed at solving research problems (for example, finding an answer to the question: “Which group used these chords for the first time ?)