def nst(
content_image: torch.Tensor,
style_image: torch.Tensor,
impl_params: bool = True,
hyper_parameters: Optional[HyperParameters] = None,
quiet: bool = False,
) -> torch.Tensor:
r"""NST from :cite:`LW2016`.
Args:
content_image: Content image for the NST.
style_image: Style image for the NST.
impl_params: Switch the behavior and hyper-parameters between the reference
implementation of the original authors and what is described in the paper.
For details see :ref:`here <li_wand_2016-impl_params>`.
hyper_parameters: If omitted,
:func:`~pystiche_papers.li_wand_2016.hyper_parameters` is used.
quiet: If ``True``, not information is logged during the optimization. Defaults
to ``False``.
"""
if hyper_parameters is None:
hyper_parameters = _hyper_parameters(impl_params=impl_params)
device = content_image.device
criterion = perceptual_loss(impl_params=impl_params,
hyper_parameters=hyper_parameters)
criterion = criterion.to(device)
image_pyramid = _image_pyramid(hyper_parameters=hyper_parameters,
resize_targets=(criterion, ))
initial_resize = image_pyramid[-1].resize_image
content_image = initial_resize(content_image)
style_image = initial_resize(style_image)
input_image = misc.get_input_image(
starting_point=hyper_parameters.nst.starting_point,
content_image=content_image)
preprocessor = _preprocessor().to(device)
postprocessor = _postprocessor().to(device)
criterion.set_content_image(preprocessor(content_image))
criterion.set_style_image(preprocessor(style_image))
return optim.pyramid_image_optimization(
input_image,
criterion,
image_pyramid,
get_optimizer=optimizer,
preprocessor=preprocessor,
postprocessor=postprocessor,
quiet=quiet,
)
def test_default_image_pyramid_optim_loop(optim_asset_loader):
asset = optim_asset_loader("default_image_pyramid_optim_loop")
actual = optim.pyramid_image_optimization(
asset.input.image,
asset.input.perceptual_loss,
asset.input.pyramid,
get_optimizer=asset.params.get_optimizer,
quiet=True,
)
desired = asset.output.image
ptu.assert_allclose(actual, desired, rtol=1e-4)
def test_default_image_pyramid_optim_loop_processing(optim_asset_loader):
asset = optim_asset_loader("default_image_pyramid_optim_loop")
actual = optim.pyramid_image_optimization(
asset.input.image,
asset.input.criterion,
asset.input.pyramid,
get_optimizer=asset.params.get_optimizer,
preprocessor=asset.params.preprocessor,
postprocessor=asset.params.postprocessor,
quiet=True,
)
desired = asset.output.image
ptu.assert_allclose(actual, desired, rtol=1e-4)
print(image_pyramid)
########################################################################################
# With a pyramid the NST is performed by
# :func:`~pystiche.optim.pyramid_image_optimization`. We time the execution and show
# the result afterwards.
#
# .. note::
#
# We regenerate the ``input_image`` since it was changed inplace during the first
# optimization.
input_image = get_input_image(starting_point, content_image=content_image)
start_with_pyramid = time.time()
output_image = optim.pyramid_image_optimization(input_image, criterion,
image_pyramid)
stop_with_pyramid = time.time()
# sphinx_gallery_thumbnail_number = 5
show_image(output_image, title="Output image with pyramid")
########################################################################################
elapsed_time_with_pyramid = stop_with_pyramid - start_with_pyramid
relative_decrease = 1.0 - elapsed_time_with_pyramid / elapsed_time_without_pyramid
print(
f"With pyramid the optimization took {elapsed_time_with_pyramid:.0f} seconds. "
f"This is a {relative_decrease:.0%} decrease.")
########################################################################################
# With the coarse-to-fine architecture of the image pyramid, the stylization of the
def guided_nst(
content_image: torch.Tensor,
content_guides: Dict[str, torch.Tensor],
style_images_and_guides: Dict[str, Tuple[torch.Tensor, torch.Tensor]],
impl_params: bool = True,
hyper_parameters: Optional[HyperParameters] = None,
quiet: bool = False,
) -> torch.Tensor:
r"""Guided NST from :cite:`GEB+2017`.
Args:
content_image: Content image for the guided NST.
content_guides: Content image guides for the guided NST.
style_images_and_guides: Dictionary with the style images and the corresponding
guides for each region.
impl_params: Switch the behavior and hyper-parameters between the reference
implementation of the original authors and what is described in the paper.
For details see :ref:`here <gatys_et_al_2017-impl_params>`.
hyper_parameters: If omitted,
:func:`~pystiche_papers.gatys_et_al_2017.hyper_parameters` is used.
quiet: If ``True``, not information is logged during the optimization. Defaults
to ``False``.
"""
regions = set(content_guides.keys())
if regions != set(style_images_and_guides.keys()):
# FIXME
raise RuntimeError
regions = sorted(regions)
if hyper_parameters is None:
hyper_parameters = _hyper_parameters()
device = content_image.device
criterion = guided_perceptual_loss(regions,
impl_params=impl_params,
hyper_parameters=hyper_parameters)
criterion = criterion.to(device)
image_pyramid = _image_pyramid(hyper_parameters=hyper_parameters,
resize_targets=(criterion, ))
initial_image_resize = image_pyramid[-1].resize_image
initial_guide_resize = image_pyramid[-1].resize_guide
content_image = initial_image_resize(content_image)
content_guides = {
region: initial_guide_resize(guide)
for region, guide in content_guides.items()
}
style_images_and_guides = {
region: (initial_image_resize(image), initial_guide_resize(guide))
for region, (image, guide) in style_images_and_guides.items()
}
input_image = misc.get_input_image(starting_point="content",
content_image=content_image)
preprocessor = _preprocessor().to(device)
postprocessor = _postprocessor().to(device)
criterion.set_content_image(preprocessor(content_image))
for region, guide in content_guides.items():
criterion.set_content_guide(guide, region=region)
for region, (image, guide) in style_images_and_guides.items():
criterion.set_style_image(preprocessor(image),
guide=guide,
region=region)
return optim.pyramid_image_optimization(
input_image,
criterion,
image_pyramid,
get_optimizer=optimizer,
preprocessor=preprocessor,
postprocessor=postprocessor,
quiet=quiet,
)
print(image_pyramid)
########################################################################################
# With a pyramid the NST is performed by
# :func:`~pystiche.optim.pyramid_image_optimization`. We time the execution and show
# the result afterwards.
#
# .. note::
#
# We regenerate the ``input_image`` since it was changed inplace during the first
# optimization.
input_image = get_input_image(starting_point, content_image=content_image)
start_with_pyramid = time.time()
output_image = optim.pyramid_image_optimization(input_image, perceptual_loss,
image_pyramid)
stop_with_pyramid = time.time()
# sphinx_gallery_thumbnail_number = 5
show_image(output_image, title="Output image with pyramid")
########################################################################################
elapsed_time_with_pyramid = stop_with_pyramid - start_with_pyramid
relative_decrease = 1.0 - elapsed_time_with_pyramid / elapsed_time_without_pyramid
print(
f"With pyramid the optimization took {elapsed_time_with_pyramid:.0f} seconds. "
f"This is a {relative_decrease:.0%} decrease.")
########################################################################################
# With the coarse-to-fine architecture of the image pyramid, the stylization of the