Audio Event Recordings For DCASE 2022 Task 3

Featuring audio files for DCASE 2022 - Task 3, sourced from the AudioSet ontology. The included labels are limited to a subset of sound events, such as female speech, male speech, clapping, and telephone sounds.

Use Cases

• Train an audio classification model to distinguish between female speech and male speech labels.
• Build a sound event detection system for clapping and telephone bell events.
• Fine-tune a model on the limited label set for DCASE 2022 Task 3 evaluation.

Strengths

• Data is curated for the specific DCASE 2022 - Task 3 challenge.
• Sound events are organized using the established AudioSet ontology hierarchy.

Limitations

• The dataset includes only a limited subset of the full AudioSet ontology labels.
• Sample data, row count, column definitions, and file formats are unknown, limiting assessment.

Provenance

Source

AudioSet ontology, subset prepared by author Fhrozen.

Collection Method

Repository includes audio files specifically for DCASE 2022 - Task 3.

Freshness

Last updated on 2023-05-07.

Geography

Region tag indicates 'us', but specific spatial coverage is unknown.