This would allow users to construct datasets from files and use them to train a model in a similar fashion to fastai's databunch construction for use with the Flux library for Julia. At the moment it only works for images but it may be updated at a later date to support text.
Download this package by running
(v1.2) pkg> add https://github.com/Sapo-Dorado/SapoDatasets
Check out the ExampleUse folder for an idea of how to use this code