Users should not have to think about how to create some kind of intermediary training dataset. They should just have to specify the dataset they want to eventually process (original source dataset) and any temporary or intermediary files should be automatically generated, without the user having to think about or manage these datasets. These could be hidden (or at least not part of standard user workflow) files in a project (for example).
See related discussion #114
Users should not have to think about how to create some kind of intermediary training dataset. They should just have to specify the dataset they want to eventually process (original source dataset) and any temporary or intermediary files should be automatically generated, without the user having to think about or manage these datasets. These could be hidden (or at least not part of standard user workflow) files in a project (for example).
See related discussion #114