Key facts
Contents & fields
The dataset contains 10 urban sound event classes: alarm-siren, children playing, dog bark, engine, footsteps, glass breaking, gun shot, metro train, rain, screams. Audio format is .wav, duration < 4 seconds, sampling rate 22KHz-44KHz. Development set: 4831 samples, evaluation set: 218 samples. Includes metadata (.csv) and source distribution per class (.png).
- alarm-siren——alarm/siren sound
- children playing——children playing sound
- dog bark——dog barking sound
- engine——engine sound
- footsteps——footsteps sound
- glass breaking——glass breaking sound
- gun shot——gun shot sound
- metro train——metro train sound
- rain——rain sound
- screams——screaming sound
Research uses
Suitable for urban sound event classification, deep learning audio recognition, environmental sound analysis, etc.
This card was drafted from the source page; institution, coverage, time span, scale, fields and license are subject to the official page (pending human review).
Keywords
Why this is hard to get on your own
When obtaining audio data for urban sound event classification, one often needs a well-labeled, balanced dataset with clear license.