gingerbread.user_package_files package#

Subpackages#

Submodules#

gingerbread.user_package_files.central_processing module#

class gingerbread.user_package_files.central_processing.CentralProcessing#

Bases: CPNeoTemplate

Central processing unit for preprocessing, postprocessing, predicting and training.

Parameters:

args (argparse.Namespace) – The arguments for the central processing unit.

Warning

Remember to include methods for preprocessing, postprocessing, predict_step or you will get an error.

postprocess(data: ndarray, extras: dict[str, Any] | None = {}) → ndarray#

Postprocess the data after training/val/test/predict

Parameters:

  • data (np.ndarray) – the data to be postprocessed (e.g image, Size [C, D, H, W]). We remove batch in the backend from predict.

  • extras (dict) –

    additional arguments for preprocessing such as resolution information etc. If provided, explain in depth in the docstring of the input and the input type. Example of extras:

    resolution [list]: resolution of the image, e.g. {“resolution”: [1.0, 1.0, 1.0]}

Important

Extras dictionary is something the researchers need to define. There has to be a proper explanation of what the extras dictionary is and what it contains, as shown in the example above.

Returns:

the postprocessed data

Return type:

np.ndarray

predict_step(data: ndarray, model: ModelInput) → ndarray#

Predict step function.

Parameters:

data (np.ndarray) – data input (e.g image, Size [C, D, H, W]) ( Hence, you need to unsqueeze the data to make it [1, C, D, H, W])

Returns:

Predictions, Must be of size [1, C, D, H, W]) the reason for this is that we might use monte carlo prediction.

Return type:

np.ndarray

preprocess(data: ndarray, extras: dict[str, Any] | None = {}) → ndarray#

Preprocess the data before training/val/test/predict.

Parameters:

  • data (np.ndarray) – the data to be preprocessed (e.g image, Size [C, D, H, W])

  • extras (dict) –

    additional arguments for preprocessing such as resolution information etc. If provided, explain in depth in the docstring of the input and the input type. Example of extras:

    resolution [list]: resolution of the image, e.g. {“resolution”: [1.0, 1.0, 1.0]}

Important

Extras dictionary is something the researchers need to define. There has to be a proper explanation of what the extras dictionary is and what it contains, as shown in the example above.

Returns:

the preprocessed data

Return type:

np.ndarray

Module contents#