AudioSet: Large-Scale Dataset of Annotated Audio Events

32.45 GB of raw audio files categorized into 632 distinct audio event classes. This collection provides the source waveforms for large-scale acoustic event detection and classification tasks.

Use Cases

• Train multi-label audio classifiers using the 632 event classes
• Perform feature engineering on 32.45 GB of raw waveforms to improve sound recognition accuracy
• Benchmark acoustic event detection models using the provided class labels

Strengths

• 32.45 GB of raw audio files
• 632 distinct audio event classes
• Raw audio format for direct waveform analysis