BIOACOUSTIC DATASETLarge-Scale Audio Event BenchmarkDOI: 10.1109/ICASSP.2017.7952261
AudioSet: An Ontology and Human-Labeled Dataset for Audio Events
Publisher / Curator: Google Research
Dataset Overview & Scope
Benchmark dataset of over 2 million 10-second audio segments drawn from YouTube videos labeled across an ontology of 527 audio event classes, including animal vocalizations and environmental sounds. Terms: Creative Commons Attribution 4.0 International (CC BY 4.0) for dataset metadata and extracted features | Access: Open web download | Download: Direct CSV annotations and 128-dimensional VGGish feature embeddings | Restrictions: YouTube terms of service apply to raw source audio video IDs.
Open Sound Repository: Official dataset repository
Access Repository DOI Permanent Identifier: Canonical citation link
Resolve DOI Keywords
audiosetgoogle-researchaudio-eventsontologybenchmarkmachine-learning