Audio Classification and Event Detection with ML Training Course

Basics of Audio Classification

• Categorization of sound events: environmental, mechanical, human-originated
• Review of practical applications: surveillance, monitoring, automation
• Distinguishing between audio classification, detection, and segmentation

Audio Data Handling and Feature Extraction

• Formats and types of audio files
• Considerations for sampling rate, windowing, and frame size
• Deriving MFCCs, chroma features, and mel-spectrograms

Data Setup and Labeling

• Utilizing UrbanSound8K, ESC-50, and bespoke datasets
• Tagging sound events and defining temporal boundaries
• Techniques for dataset balancing and audio augmentation

Developing Audio Classification Models

• Application of convolutional neural networks (CNNs) to audio data
• Input formats: raw waveform versus extracted features
• Loss functions, evaluation criteria, and managing overfitting

Event Detection and Time-based Localization

• Detection strategies based on frames and segments
• Refining detection results through thresholding and smoothing
• Mapping predictions onto audio timelines for visualization

Advanced Concepts and Live Processing

• Utilizing transfer learning in data-scarce situations
• Model deployment using TensorFlow Lite or ONNX
• Handling streaming audio and managing latency

Project Construction and Use Cases

• Architecting a complete pipeline from ingestion to classification
• Creating proof-of-concept solutions for surveillance, quality control, or monitoring
• Implementing logging, alerts, and integration with dashboards or APIs

Conclusion and Future Directions