morpheus_framework package

Subpackages

Submodules

morpheus_framework.morpheus_framework module

class morpheus_core.morpheus_core.AGGREGATION_METHODS[source]

Bases: object

Helper class with string constants to use as arguments in morpheus_core methods.

INVALID_ERR = 'Invalid aggregation method please select one of AGGREGATION_METHODS.MEAN_VAR or AGGREGATION_METHODS.RANK_VOTE'
MEAN_VAR = 'mean_var'
RANK_VOTE = 'rank_vote'
morpheus_core.morpheus_core.predict(model: Callable, model_inputs: List[Union[str, numpy.ndarray]], n_classes: int, batch_size: int, window_shape: Tuple[int, int], dilation: float = 1, stride: Tuple[int, int] = (1, 1), update_map: numpy.ndarray = None, aggregate_method: str = 'rank_vote', out_dir: str = None, gpus: List[int] = None, cpus: int = None, parallel_check_interval: float = 1) → Tuple[List[astropy.io.fits.hdu.hdulist.HDUList], List[numpy.ndarray]][source]

Applies the model the model_inputs

If you are using the parallel functionality, then model must be pickleable.

Parameters:
  • model (Callable) – The model to apply to the inputs
  • model_inputs (List[Union[np.ndarray, str]]) – The inputs to classify using the given model
  • n_classes (int) – The number of classes that are output
  • batch_size (int) – The number of examples to include in a batch
  • window_shape (int) – The (height, width) of the samples to extract
  • stride (Tuple[int, int]) – How many (rows, columns) to move through the image at each iteration.
  • update_map (np.narray) – A 2D array of the same size as window height that indicates which pixels to use to updates for each example
  • aggregate_method (str) – How to process the output from the model. If AGGREGATION_METHODS.MEAN_VAR record output using mean and variance, If AGGREGATION_METHODS.RANK_VOTE record output as the normalized vote count.
  • out_dir (str) – The directory to save output files in if the model_inputs are string locations.
  • gpus (List[int]) – The gpu ids to use for parallel processesing
  • cpus (int) – The number of cpus to use for parllel processing
Returns:

A 2-Tuple where the first element is the list of fits.HDULS for the outputfiles. The second element is a list of the output arrays from the model given the the input arrays.
Raises:
  • ValueError if model_inputs are not all of the same type
  • ValueError if model_inputs are not str or np.ndarray
  • ValueError if both gpus and cpus are given
  • ValueError is cpus or gpus are given, but out_dir is not given
  • ValueError if len(gpus)==1
  • ValueError if cpus<2

Module contents