Skip to content

semipy.datasets.utils.build_augmentations

Warning

This section is in construction.

    semipy.datasets.utils.build_augmentations(dset_name: str,
                                              path: Optional[str] = None,
                                              transforms: Dict = None,
                                              use_strong_augmentation: bool = True) -> Tuple[torchvision.transforms.Compose, torchvision.transforms.Compose]

This functions reads the TRANSFORMS part from DATA in the configuration file in order to create two compositions of transformations (usually a weak and a strong one). If nothing is specified in this part of the configuration file, a list of default transformations will be used.

Parameters

  • dset_name (str) - Name of the used dataset. Use 'CUSTOM' for a custom one.
  • path (optional, str) - Path to a YAML file containing a list of transformations. If None, the default augmentations.yaml file will be used. Default: None
  • transforms (dict) - Dictionary containing both 'WEAK' and 'STRONG' lists of transformations. Default: None
  • use_strong_augmentation (bool) - Toggle the usage of the second list of transformations (usually the strong augmentation). Default: True