pyraug.data¶
This module contrains the methods to load and preprocess the data.
Note
As of now, only imaging modality is handled by Pyraug. In the near future other modalities should be added.
Datasets¶
The pyraug’s Datasets inherit from
torch.utils.data.Dataset
and must be used to convert the data before
training. As of today, it only contains the pyraug.data.BaseDatset
useful to train a
VAE model but other Datatsets will be added as models are added.
This class is the Base class for pyraug’s dataset |
Loaders¶
The loaders are used to load the data from a particular format to numpy.ndarray
or
List[numpy.ndarray]
This is the Base data loader from which all future loaders must inherit. |
|
This loader allows you to load imagining data directly from a folder and convert it to |
Preprocessors¶
The purpose of the preprocessor is to ensure the data is not corrupted (no nan), reshape
it in case inconsistencies are detected, normalize it and converted it to a format handled by the
Trainer
. In particular, an input data is converted to a
torch.Tensor
and all the data is gather into a BaseDatset
instance.
By choice, we do not provided very advanced preprocessing functions (such as image registrations) since the augmentation method should be robust to huge differences in the data and be able to reproduce and account for this diversity. More advanced preprocessing is up to the user.
This is a basic class which preprocesses the data. |