GiantMIDI-Piano

dataset·machine transcribed piano midi·active

Machine-transcribed solo-piano MIDI derived from web recordings and IMSLP work metadata.

Recorded facts

Official site	https://github.com/bytedance/GiantMIDI-Piano ↗
Geography	Global
arxiv	2010.07061
github repository	bytedance/GiantMIDI-Piano
size	works: 10855 · composers: 2786 · curated works: 7236 · transcribed notes: 38700838
scope	Large-scale classical piano symbolic data for analysis and generation.
genres	classical piano
steward	ByteDance research authors
creators	Qiuqiang Kong; Bochen Li; Jitong Chen; Yuxuan Wang
languages	multilingual titles
modalities	MIDI transcriptions; work and composer metadata; source-retrieval code
geographies	Global
access method	GitHub code and generated MIDI resources.
intended uses	symbolic music generation; classical music analysis; piano transcription research
related papers	GiantMIDI-Piano: A large-scale MIDI dataset for classical piano music
annotation method	CNN solo-piano detection followed by automatic high-resolution transcription.
collection method	Work names from IMSLP used to search and download corresponding internet audio recordings.
provenance claims	Repository releases code for acquisition; it does not redistribute one clearly licensed audio corpus.
current availability	Official repository available.
us market scope basis	Publicly available or materially used in US-facing music/audio-AI research.
limitations biases disputes	MIDI is machine transcribed and contains detection, metadata and transcription errors.; Web recording rights are not resolved by source code release.