Dataset

This module holds all of the pytorch dataset classes

niftidataset.dataset

the actual dataset classes of niftidataset

Author: Jacob Reinhold (jacob.reinhold@jhu.edu)

Created on: Oct 24, 2018

niftidataset.dataset.train_val_split(source_dir: str, target_dir: str, valid_pct: float = 0.2, transform: Optional[Callable] = None, preload: bool = False)

create two separate NiftiDatasets in PyTorch for working with NifTi files. If a directory contains source files and the other one contains target files and also you dont have a specific directory for validation set, this function splits data to two NiftiDatasets randomly with given percentage.

Parameters:

• source_dir (str) – path to source images.
• target_dir (str) – path to target images.
• valid_pct (float) – percent of validation set from data.
• transform (Callable) – transform to apply to both source and target images.
• preload – load all data when initializing the dataset

Returns:

(train_dataset, validation_dataset).

Return type:

Tuple