We looked at a completely new problem and case study in this chapter, involving audio identification and classification. Concepts surrounding audio data and signals were covered, including effective techniques to visualize and understand this data type.

We also looked at effective feature engineering techniques and how we could use transfer learning to extract effective features from image representations of audio data. This shows us the promise of transfer learning and how you can leverage knowledge from one domain (images) to another domain (audio) and build an extremely robust and effective classifier. Finally, we built a complete end-to-end pipeline for identifying and classifying new samples of audio data. Do check out further datasets of annotated audio on the web, and see if you can build a bigger and better audio identifier and classifier by leveraging concepts from transfer learning that you learned here. Stay tuned for more interesting examples and case studies on transfer learning.