OODDataset: Abstraction to manage dataset and provide methods suited to OOD detection problems, like assigning OOD label to datasets, concatenating and filtering by labels. Relies of DataHandler to perform library dependent actions.
TFDataHandling: low level class to manage tf.data.Datasets underlying OODDatasets.
The idea is to encapsulate library dependent syntax as much as possible inside TFDataHandler to eventually prepare a TorchDataHandler (Left for future developments), and to provide a high level, readable API for OODDatasets.
Creates two new classes:
The idea is to encapsulate library dependent syntax as much as possible inside TFDataHandler to eventually prepare a TorchDataHandler (Left for future developments), and to provide a high level, readable API for OODDatasets.