jpalpant / audioanalysis Goto Github PK

Version 0.1 implemented no sane class structure - the first step in developing v0.2 will be to define a sane class hierarchy and set of interfaces for the next several stages of the project, keeping in mind the focus on classification and audio data.

Note that this hierarchy does not need to consider the GUI at all - that will be developed later to fit the library, not the other way around.

v0.1.0 demoed a convolutional neural network which operated in the image space of the spectrogram of the audio data. While the method is valuable (and should be expanded), it is not the only option. Several other classifiers should be explored, developed and tested.

Viable candidates will be recorded here - development can't begin until a class hierarchy is developed.

Create a complete set of documentation for the package developed for version 0.2.0 using Sphinx and host it on Read the Docs.

Version 0.1.0 implemented no unit testing whatsoever - a situation that can't last in the long term. During the course of the development of the new project structure, implement unit testing and CI using a cloud-based CI provider of your choice. I'm leaning towards Travis, on account of rnaseqflow using Circle.

Since separation of GUI and library are to be enforced and tests implemented, the shape of the repository is going to change. Define a new repository structure that is conducive to further development.