DEAM

dataset·music emotion·active

Music excerpts and full songs with continuous and song-level valence and arousal annotations.

Recorded facts

Official site	https://cvml.unige.ch/databases/DEAM/ ↗
Geography	Europe
size	songs and excerpts: 1802
scope	Dynamic music emotion recognition.
genres	multiple
license	Creative Commons audio with per-item terms
steward	University of Geneva CVML and MediaEval collaborators
creators	Anna Aljanaki; Yi-Hsuan Yang; Mohammad Soleymani
languages	multiple/unknown
modalities	Creative Commons audio; continuous valence/arousal; song-level valence/arousal; openSMILE features; metadata
geographies	Global
access method	Direct audio, annotations and feature downloads.
intended uses	music emotion recognition; affective computing
related papers	Developing a benchmark for emotional analysis of music
annotation method	Listeners continuously and globally rated perceived valence and arousal.
collection method	Creative Commons music assembled for MediaEval emotion tasks.
current availability	Official downloads available.
commercial use limits	Depends on source-track license.
us market scope basis	Publicly available or materially used in US-facing music/audio-AI research.
limitations biases disputes	Perceived emotion is subjective and culturally dependent.; Genre and source distributions are uneven.
redistribution restrictions	Per-track Creative Commons terms apply.
related models tools benchmarks	MediaEval Emotion in Music