TAU Urban Acoustic Scenes 2019: 10-Second Audio Segments from 10 Urban Environments

TAU Urban Acoustic Scenes 2019 evaluation dataset consists of 10-second audio segments from 10 distinct acoustic scenes in urban environments. The dataset contains a total of 20 hours of audio. It was created by Toni Heittola of Tampere University.

Use Cases

• Train machine learning models for acoustic scene classification based on the 10 labeled urban scene categories.
• Benchmark audio classification algorithms using a standardized evaluation dataset of urban sounds.
• Study the acoustic characteristics of different urban environments like metro stations, parks, and pedestrian streets.

Strengths

• Contains 20 hours of total audio data, providing a substantial corpus for analysis.
• Features 10 distinct and clearly defined urban acoustic scene categories, such as airport, metro station, and park.
• Audio segments are standardized to a 10-second duration, which likely facilitates consistent model input.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Last update date is unknown; freshness unverified.
• The specific geographic origin of the audio recordings is not stated, which may limit generalizability.

Provenance

Source

Tampere University