dasf.transforms.operations
==========================

.. py:module:: dasf.transforms.operations

.. autoapi-nested-parse::

   Basic transform operations module. 



Classes
-------

.. autoapisummary::

   dasf.transforms.operations.Reshape
   dasf.transforms.operations.SliceArray
   dasf.transforms.operations.SliceArrayByPercent
   dasf.transforms.operations.SliceArrayByPercentile
   dasf.transforms.operations.ApplyPatchesBase
   dasf.transforms.operations.ApplyPatchesWeightedAvg
   dasf.transforms.operations.ApplyPatchesVoting


Module Contents
---------------

.. py:class:: Reshape(shape = None)

   Bases: :py:obj:`dasf.transforms.base.Fit`


   Get a slice of a cube. An inline slice is a section over the x-axis.

   Parameters
   ----------
   iline_index : int
       The index of the inline to get.



   .. py:attribute:: shape


   .. py:method:: fit(X, y=None)

      Generic fit funtion according executor.



.. py:class:: SliceArray(output_size)

   Bases: :py:obj:`dasf.transforms.base.Transform`


   Class representing a Transform operation of the pipeline.


   .. py:attribute:: x


   .. py:method:: transform(X)

      Generic transform funtion according executor.



.. py:class:: SliceArrayByPercent(x=100.0, y=100.0, z=100.0)

   Bases: :py:obj:`dasf.transforms.base.Transform`


   Class representing a Transform operation of the pipeline.


   .. py:attribute:: x


   .. py:attribute:: y


   .. py:attribute:: z


   .. py:method:: transform(X)

      Generic transform funtion according executor.



.. py:class:: SliceArrayByPercentile(percentile)

   Bases: :py:obj:`dasf.transforms.base.Transform`


   Class representing a Transform operation of the pipeline.


   .. py:attribute:: p


   .. py:method:: _internal_chunk_array_positive(block, axis=None, keepdims=False, xp=np)


   .. py:method:: _internal_aggregate_array_positive(block, axis=None, keepdims=False, xp=np)


   .. py:method:: _internal_chunk_array_negative(block, axis=None, keepdims=False, xp=np)


   .. py:method:: _internal_aggregate_array_negative(block, axis=None, keepdims=False, xp=np)


   .. py:method:: _lazy_transform_cpu(X)

      Respective lazy transform mocked function for CPUs.



   .. py:method:: _lazy_transform_gpu(X)

      Respective lazy transform mocked function for GPUs.



   .. py:method:: _transform_cpu(X)

      Respective immediate transform mocked function for local CPU(s).



   .. py:method:: _transform_gpu(X)

      Respective immediate transform mocked function for local GPU(s).



.. py:class:: ApplyPatchesBase(function, weight_function, input_size, overlap, offsets)

   Bases: :py:obj:`dasf.transforms.base.Transform`


   Base Class for ApplyPatches Functionalities

   function: function to be applied to each patch, can be eiter a Python Function or a ModelLoader
   weight_function: weight attribution function, must receive a shape and produce a NDArray with the respective weights for each array position
   input_size: size of input to the function to be applied,
   overlap: dictionary containing overlapping/padding configurations to use with np.pad or dask.overlap.overlap. Its important that for the base patch set the whole "chunk core" is covered by the patches.
   offsets: list of offsets for overlapping patches extraction


   .. py:attribute:: _function


   .. py:attribute:: _weight_function


   .. py:attribute:: _input_size


   .. py:attribute:: _offsets


   .. py:attribute:: overlap


   .. py:attribute:: _overlap_config


   .. py:method:: _apply_patches(patch_set)

      Applies function to each patch in a patch set




   .. py:method:: _reconstruct_patches(patches, index, weights, inner_dim=None)

      Rearranges patches to reconstruct area of interest from patches and weights



   .. py:method:: _adjust_patches(arrays, ref_shape, offset, pad_value=0)

      Pads reconstructed_patches with 0s to have same shape as the reference shape from the base patch set



   .. py:method:: _combine_patches(results, offsets, indexes)
      :abstractmethod:


      How results are combined is dependent on what is being combined.
      ApplyPatchesWeightedAvg uses Weighted Average
      ApplyPatchesVoting uses Voting (hard or soft)



   .. py:method:: _extract_patches(data, patch_shape)

      Patch extraction method. It will be called once for the base patch set and also for the requested offsets (overlapping patch sets)



   .. py:method:: _operation(chunk)

      Operation to be performed on each chunk



   .. py:method:: _transform(X)


   .. py:method:: _lazy_transform(X)


   .. py:method:: _lazy_transform_cpu(X, **kwargs)

      Respective lazy transform mocked function for CPUs.



   .. py:method:: _lazy_transform_gpu(X, **kwargs)

      Respective lazy transform mocked function for GPUs.



   .. py:method:: _transform_cpu(X, **kwargs)

      Respective immediate transform mocked function for local CPU(s).



   .. py:method:: _transform_gpu(X, **kwargs)

      Respective immediate transform mocked function for local GPU(s).



.. py:class:: ApplyPatchesWeightedAvg(function, weight_function, input_size, overlap, offsets)

   Bases: :py:obj:`ApplyPatchesBase`


   ApplyPatches with Weighted Average combination function.

   function: function to be applied to each patch, can be eiter a Python Function or a ModelLoader
   weight_function: weight attribution function, must receive a shape and produce a NDArray with the respective weights for each array position
   input_size: size of input to the function to be applied,
   overlap: dictionary containing overlapping/padding configurations to use with np.pad or dask.overlap.overlap. Its important that for the base patch set the whole "chunk core" is covered by the patches.
   offsets: list of offsets for overlapping patches extraction


   .. py:method:: _combine_patches(results, offsets, indexes)

      How results are combined is dependent on what is being combined.
      ApplyPatchesWeightedAvg uses Weighted Average
      ApplyPatchesVoting uses Voting (hard or soft)



.. py:class:: ApplyPatchesVoting(function, weight_function, input_size, overlap, offsets, voting, num_classes)

   Bases: :py:obj:`ApplyPatchesBase`


   ApplyPatches with Voting combination function.

   function: function to be applied to each patch, can be eiter a Python Function or a ModelLoader
   weight_function: weight attribution function, must receive a shape and produce a NDArray with the respective weights for each array position
   input_size: size of input to the function to be applied,
   overlap: dictionary containing overlapping/padding configurations to use with np.pad or dask.overlap.overlap. Its important that for the base patch set the whole "chunk core" is covered by the patches.
   offsets: list of offsets for overlapping patches extraction
   voting: voting method. "hard"  or "soft"
   num_classes: number of classes possible


   .. py:attribute:: _voting


   .. py:attribute:: _num_classes


   .. py:method:: _combine_patches(results, offsets, indexes)

      How results are combined is dependent on what is being combined.
      ApplyPatchesWeightedAvg uses Weighted Average
      ApplyPatchesVoting uses Voting (hard or soft)



   .. py:method:: _hard_voting(results, offsets, indexes)

      Hard voting combination function



   .. py:method:: _soft_voting(results, offsets, indexes)

      Soft voting combination function