iirc package

Subpackages

• iirc.lifelong_dataset package
  • Submodules
  • iirc.lifelong_dataset.base_dataset module
  • iirc.lifelong_dataset.tensorflow_dataset module
  • iirc.lifelong_dataset.torch_dataset module
  • Module contents
• iirc.utils package
  • Submodules
  • iirc.utils.download_cifar module
  • iirc.utils.prepare_imagenet module
  • iirc.utils.utils module
  • Module contents

Submodules

iirc.datasets_loader module

iirc.datasets_loader.get_lifelong_datasets(dataset_name: str, dataset_root: str = './data', setup: str = 'IIRC', framework: str = 'PyTorch', tasks_configuration_id: int = 0, essential_transforms_fn: Optional[Callable[[PIL.Image.Image], Any]] = None, augmentation_transforms_fn: Optional[Callable[[PIL.Image.Image], Any]] = None, cache_images: bool = False, joint: bool = False) → Tuple[Dict[str, iirc.lifelong_dataset.base_dataset.BaseDataset], List[List[str]], Dict[str, int]]

Get the incremental refinement learning , as well as the tasks (which contains the classes introduced at each task), and the index for each class corresponding to its order of appearance

Parameters

• dataset_name (str) – The name of the dataset, ex: iirc_cifar100

• dataset_root (str) – The directory where the dataset is/will be downloaded (default: “./data”)

• setup (str) – Class Incremental Learning setup (CIL) or Incremental Implicitly Refined Classification setup (IIRC) (default: IIRC_SETUP)

• framework (str) – The framework to be used, whether PyTorch or Tensorflow. use Tensorflow for any numpy based dataloading (default: PYTORCH)

• tasks_configuration_id (int) – The configuration id, where each configuration corresponds to a specific tasks and classes order for each dataset. This id starts from 0 for each dataset. Ignore when joint is set to True (default: 0)

• essential_transforms_fn (Optional[Callable[[Image.Image], Any]]) – A function that contains the essential transforms (for example, converting a pillow image to a tensor) that should be applied to each image. This function is applied only when the augmentation_transforms_fn is set to None (as in the case of a test set) or inside the disable_augmentations context (default: None)

• augmentation_transforms_fn – A function that contains the essential transforms (for example, converting a pillow image to a tensor) and augmentation transforms (for example, applying random cropping) that should be applied to each image. When this function is provided, essential_transforms_fn is not used except inside the disable_augmentations context (default: None)

• cache_images (bool) – cache images that belong to the current task in the memory, only applicable when using the image path (default: False)

• joint (bool) – provided all the classes in a single task for joint training (default: False)

Returns

lifelong_datasets (Dict[str, BaseDataset]): a dictionary with the keys corresponding to the four splits (train, intask_validation, posttask_validation, test), and the values containing the dataset object inheriting from BaseDataset for that split.

tasks (List[List[str]]): a list of lists where each inner list contains the set of classes (class names) that will be introduced in that task (example: [[dog, cat, car], [tiger, truck, fish]]).

class_names_to_idx (Dict[str, int]): a dictionary with the class name as key, and the class index as value (example: {“dog”: 0, “cat”: 1, …}).

Return type

Tuple[Dict[str, BaseDataset], List[List[str]], Dict[str, int]]

iirc.definitions module

Module contents