UrbanSound8K Audio Classification Dataset

UrbanSound8K is an audio dataset containing 8,732 annotated urban sound clips (≤4 seconds). The dataset contains 27 hours of audio, of which 18.5 hours of audio are annotated with event occurrences.

The audios fall into 10 categories: air conditioner, car horn, children playing, dog barking, drilling, engine sound, gunshot, hammer, whistle and street music. The categories are based on the urban sound taxonomy. All the sound clips are from field recordings uploaded to www.freesound.org.