SALAMI

dataset·music structure annotations·active

Hierarchical expert structural annotations for a large and stylistically diverse music collection.

Verified facts

Official site	https://github.com/DDMAL/salami-data-public ↗
Geography	Canada
github repository	DDMAL/salami-data-public
size	public tracks: 1359 · two annotator tracks: 884
scope	Music structure analysis and evaluation.
genres	multiple
steward	McGill DDMAL and SALAMI collaborators
creators	Jordan B. L. Smith; John Ashley Burgoyne; Ichiro Fujinaga; David De Roure; J. Stephen Downie
languages	not_applicable
modalities	hierarchical segment boundaries; segment labels; metadata
geographies	Global
access method	Public annotation repository; users source audio separately.
intended uses	music structure analysis; segmentation benchmarking
related papers	Design and creation of a large-scale database of structural annotations
annotation method	Music experts produced hierarchical structural annotations, often with two annotators per track.
collection method	Music selected across collections; audio is not distributed in the public annotations.
current availability	Public annotations available.
us market scope basis	Publicly available or materially used in US-facing music/audio-AI research.
limitations biases disputes	Audio is not distributed.; Annotators can disagree on hierarchical boundaries and labels.
related models tools benchmarks	McGill Billboard Project; mirdata