def apply_transform(self, sample: Subject) -> dict:
bspline_params = self.get_params(
self.num_control_points,
self.max_displacement,
self.num_locked_borders,
)
random_parameters_dict = {'coarse_grid': bspline_params}
if not self.is_tensor:
sample.check_consistent_shape()
for image_dict in sample.get_images(intensity_only=False):
if image_dict[TYPE] == LABEL:
interpolation = Interpolation.NEAREST
else:
interpolation = self.interpolation
image_dict[DATA] = self.apply_bspline_transform(
image_dict[DATA],
image_dict[AFFINE],
bspline_params,
interpolation,
)
sample.add_transform(self, random_parameters_dict)
else:
sample = self.apply_bspline_transform(
sample,
np.identity(4),
bspline_params,
self.interpolation
)
return sample
def apply_transform(self, sample: Subject) -> dict:
scaling_params, rotation_params = self.get_params(
self.scales, self.degrees, self.isotropic)
random_parameters_dict = {
'scaling': scaling_params,
'rotation': rotation_params,
}
if not self.is_tensor:
sample.check_consistent_shape()
for image_dict in sample.get_images(intensity_only=False):
if image_dict[TYPE] == LABEL:
interpolation = Interpolation.NEAREST
else:
interpolation = self.interpolation
image_dict[DATA] = self.apply_affine_transform(
image_dict[DATA],
image_dict[AFFINE],
scaling_params,
rotation_params,
interpolation,
)
sample.add_transform(self, random_parameters_dict)
else:
sample = self.apply_affine_transform(
sample,
np.identity(4),
scaling_params,
rotation_params,
self.interpolation,
)
return sample
def apply_transform(self, sample: Subject) -> dict:
axes_to_flip_hot = self.get_params(self.axes, self.flip_probability)
random_parameters_dict = {'axes': axes_to_flip_hot}
if self.is_tensor:
return self.flip_dimensions(sample, axes_to_flip_hot)
else:
for image_dict in sample.get_images(intensity_only=False):
tensor = image_dict[DATA]
image_dict[DATA] = self.flip_dimensions(tensor, axes_to_flip_hot)
sample.add_transform(self, random_parameters_dict)
return sample
def apply_transform(self, sample: Subject) -> dict:
random_parameters_images_dict = {}
std = self.get_params(self.std_range)
if not self.is_tensor:
sample.check_consistent_shape()
for image_name, image_dict in sample.get_images_dict().items():
random_parameters_dict = {'std': std}
random_parameters_images_dict[image_name] = random_parameters_dict
image_dict[DATA] = add_noise(image_dict[DATA], std)
sample.add_transform(self, random_parameters_images_dict)
else:
sample = add_noise(sample, std)
return sample