Explore the powerful Python library "praudio" for batch preprocessing audio datasets in a single command. Learn about the design and implementation of this tool, which leverages Librosa as its processing backend. Discover the codebase structure, including IO facilities, Signal objects, and the design of transforms and preprocessors. Understand how to create praudio objects, install the library from PyPi or GitHub, and utilize it in your audio AI projects. Dive into preprocessing configurations, the preprocess script, and see real-world results of running the tool. Gain insights into the library's current limitations and potential future improvements to enhance your audio data preprocessing workflows.
Overview
Syllabus
Intro
What does praudio do?
Librosa as processing backend
Codebase structure
IO facilities and Signal object
Transforms design
Transforms chain implementation
What do preprocessors do?
FilePreprocessor implementation
BatchFilePreprocessor design
Creation of praudio objects
Installing praudio PyPi and GitHub
Two ways to use praudio
Understanding preprocessing configurations
The preprocess script
Running preprocess + results
Limitations and future improvements
Taught by
Valerio Velardo - The Sound of AI
