pystiche_papers.li_wand_2016.rst

pystiche_papers.li_wand_2016

Title	Combining Markov Random Fields and Convolutional Neural Networks for Image Synthesis
Authors	Chuan Li and Michael Wand
Citation	:cite:`LW2016`
Reference implementation	Repository / Archive
Variant	Image optimization
Content loss	:class:`~pystiche.loss.FeatureReconstructionLoss`
Style loss	:class:`~pystiche.loss.MRFLoss`
Regularization	:class:`~pystiche.loss.TotalVariationLoss`

Behavioral changes

.. seealso::
  :ref:`Paper implementations <impl_params>`

The following parts are affected:

• :class:`~pystiche_papers.li_wand_2016.FeatureReconstructionLoss`
• :class:`~pystiche_papers.li_wand_2016.MRFLoss`
• :class:`~pystiche_papers.li_wand_2016.TotalVariationLoss`
• :class:`~pystiche_papers.li_wand_2016.target_transforms`

Hyper parameters

.. seealso::
  :ref:`Paper implementations <impl_params>`

:func:`~pystiche_papers.li_wand_2016.content_loss`

Parameter	impl_params
	True	False
layer	"relu4_1"	"relu_4_2"
score_weight	2e1	1e0

:func:`~pystiche_papers.li_wand_2016.target_transforms`

Parameter	impl_params
	True	False
num_scale_steps	0	3
scale_step_width	5e-2
num_rotate_steps	0	2
rotate_step_width	7.5

:func:`~pystiche_papers.li_wand_2016.style_loss`

Parameter	impl_params
	True	False
layers	("relu3_1", "relu4_1")
layer_weights	"sum"
patch_size	3
stride	2	1
score_weight	1e-4	1e0

:func:`~pystiche_papers.li_wand_2016.regularization`

Parameter	impl_params
	True	False
score_weight	1e-3

:func:`~pystiche_papers.li_wand_2016.image_pyramid`

Parameter	impl_params
	True	False
max_edge_size	384
num_steps	100	200
num_levels	3	None [1]
min_edge_size	64
edge	"long"

[1]	num_levels=None implies that the number of levels is automatically calculated depending on max_edge_size and min_edge_size. See :class:`pystiche.pyramid.OctaveImagePyramid` for details.

API

.. automodule:: pystiche_papers.li_wand_2016

.. autofunction:: images

.. autoclass:: FeatureReconstructionLoss

.. autofunction:: content_loss

.. autoclass:: MRFLoss

.. autofunction:: style_loss

.. autoclass:: TotalVariationLoss

.. autofunction:: regularization

.. autofunction:: perceptual_loss

.. autofunction:: nst

.. autofunction:: image_pyramid

.. autofunction:: hyper_parameters

.. autofunction:: extract_normalized_patches2d

.. autofunction:: target_transforms

.. autofunction:: preprocessor

.. autofunction:: postprocessor

.. autofunction:: multi_layer_encoder

.. autofunction:: optimizer