Package Overview

AVA is composed of five subpackages: segmenting, preprocessing, models, data, and plotting. There are three sorts of analyses AVA can do: a syllable-level analysis, a shotgun VAE analysis, and a warped-time shotgun VAE analysis, and each of these make use of the five subpackages in this order:

• The segmenting subpackage identifies syllables for syllable-level analysis and song renditions for the warped-time shotgun VAE analysis.
• The preprocessing subpackage creates spectrograms to feed into the VAE.
• The models subpackage defines the VAE as well as PyTorch DataLoader objects that feed data to the VAE.
• The data subpackage defines the DataContainer object, which is useful for organizing various files for subsequent plotting and analysis.
• The plotting subpackage contains various functions for plotting and analysis.

The remaining sections go through each subpackage in more detail.

If you prefer to learn from examples, see the examples/ subdirectory in the code, which contains python scripts implementing these three sorts of analysis.