TAU Urban Acoustic Scenes 2020 3Class: Audio for Scene Classification

TAU Urban Acoustic Scenes 2020 3Class development dataset provides 10-second audio segments from 10 specific urban environments, grouped into three major classes: indoor, outdoor, and transportation. Its structure is designed for benchmarking machine learning models in acoustic scene classification. The dataset's cross-platform presence on Papers with Code indicates its established use in the research community.

Use Cases

• [Acoustic scene classification] based on [10-second audio segments from 10 urban scenes]
• [Model benchmarking for audio AI] based on [a structured three-class hierarchy of indoor, outdoor, and transportation sounds]
• [Urban soundscape analysis] based on [recordings from environments like airports, metro stations, and public squares]

Strengths

• Provides a clear, hierarchical classification structure with 10 specific scenes grouped into 3 major classes.
• Contains standardized 10-second audio segments, facilitating consistent model training and evaluation.
• Has an Open Access (green) license, promoting reuse and dissemination in research.

Limitations

• Key metadata such as row count, total size, and specific column/feature names are unavailable across all sources.
• The dataset's freshness is unclear as no last updated date is provided.
• Descriptions are repetitive across sources, suggesting sparse documentation beyond the core classification scheme.

Provenance

Source

Tampere University

Geography

Urban environments