estimators

Estimator

Bases: Generic[ModuleT, OptionalParameterizationT, OptionalMaskerT], Module

Defines an interface for estimating noise components from input values.

Estimators should return a tensor of a broadcastable shape to the input tensor.

Added in version 0.37.0.

__call__

__call__(input: Tensor, noise_mask: Tensor | None = None, **kwargs: Any) -> tuple[torch.Tensor, torch.Tensor | None]

Estimate noise components from input values.

Parameters:

Name	Type	Description	Default
input	Tensor	The tensor to estimate noise components from.	required
noise_mask	Tensor | None	An optional mask tensor to use to select a subset of the elements of the estimated standard deviations for computing a mask to apply to the input. If None, the mask is computed over all of the estimated standard deviations.	None
kwargs	Any	Additional keyword arguments to module.	required

Returns:

Type	Description
tuple[torch.Tensor, torch.Tensor | None]	A tuple containing the estimated noise components and an input mask if there is a masker.

__init__

__init__(module: ModuleT, target_parameter: str, parameterization: OptionalParameterizationT = None, masker: OptionalMaskerT = None) -> None

Construct an Estimator with the given parameters.

Parameters:

Name	Type	Description	Default
module	ModuleT	The module to use to estimate the unparameterized noise component.	required
target_parameter	str	The name of the parameter to pass the input tensor as to the module.	required
parameterization	OptionalParameterizationT	The optional parameterization to apply to the output of module.	None
masker	OptionalMaskerT	The optional masker to apply to the output of the parameterization.	None

forward

forward(input: Tensor, noise_mask: Tensor | None = None, **kwargs: Any) -> tuple[torch.Tensor, torch.Tensor | None]

Estimate noise components from input values.

Parameters:

Name	Type	Description	Default
input	Tensor	The tensor to estimate noise components from.	required
noise_mask	Tensor | None	An optional mask tensor to use to select a subset of the elements of the estimated standard deviations for computing a mask to apply to the input. If None, the mask is computed over all of the estimated standard deviations.	None
kwargs	Any	Additional keyword arguments to module.	required

Returns:

Type	Description
tuple[torch.Tensor, torch.Tensor | None]	A tuple containing the estimated noise components and an input mask if there is a masker.

PatchEstimator

Bases: Estimator[Conv2d, OptionalParameterizationT, OptionalMaskerT]

Defines an interface for estimating noise components using non-overlapping 2D convolutions over input image(s).

Added in version 0.40.0.

__call__

__call__(input: Tensor, noise_mask: Tensor | None = None, **kwargs: Any) -> tuple[torch.Tensor, torch.Tensor | None]

Estimate noise components from input values.

Parameters:

Name	Type	Description	Default
input	Tensor	The tensor to estimate noise components from.	required
noise_mask	Tensor | None	An optional mask tensor to use to select a subset of the elements of the estimated standard deviations for computing a mask to apply to the input. If None, the mask is computed over all of the estimated standard deviations.	None
kwargs	Any	Additional keyword arguments to module.	required

Returns:

Type	Description
tuple[torch.Tensor, torch.Tensor | None]	A tuple containing the estimated noise components and an input mask if there is a masker.

__init__

__init__(module: Conv2d, target_parameter: str = 'input', parameterization: OptionalParameterizationT = None, masker: OptionalMaskerT = None, padding_mode: Literal['constant', 'reflect', 'replicate', 'circular'] = 'constant', padding_value: float = 0.0) -> None

Construct a PatchEstimator with the given parameters.

Parameters:

Name	Type	Description	Default
module	Conv2d	The 2D convolutional module to use. The estimator must have a stride equal to its kernel size (non-overlapping) and output channels equal to the input channels times the product of the kernel side lengths (input volume = output volume).	required
target_parameter	str	The name of the parameter to pass the input tensor as to the module.	'input'
parameterization	OptionalParameterizationT	The optional parameterization to apply to the output of the estimator.	None
masker	OptionalMaskerT	The optional masker to apply to the output of the parameterization.	None
padding_mode	Literal['constant', 'reflect', 'replicate', 'circular']	Type of padding. One of: constant, reflect, replicate, or circular.	'constant'
padding_value	float	Fill value for constant padding.	0.0

Raises:

Type	Description
ValueError	If the stride is not equal to the kernel size.
ValueError	If the output channels are not equal to the input channels times the product of the kernel side lengths.
ValueError	If a non-zero padding value is given for a padding mode other than constant.

forward

forward(input: Tensor, noise_mask: Tensor | None = None, **kwargs: Any) -> tuple[torch.Tensor, torch.Tensor | None]

Estimate noise components from input values.

Parameters:

Name	Type	Description	Default
input	Tensor	The tensor to estimate noise components from.	required
noise_mask	Tensor | None	An optional mask tensor to use to select a subset of the elements of the estimated standard deviations for computing a mask to apply to the input. If None, the mask is computed over all of the estimated standard deviations.	None
kwargs	Any	Additional keyword arguments to module.	required

Returns:

Type	Description
tuple[torch.Tensor, torch.Tensor | None]	A tuple containing the estimated noise components and an input mask if there is a masker.

module_from_input_shape staticmethod

module_from_input_shape(color_channels: int, patch_size: int | tuple[int, int] | Sequence[int]) -> nn.Conv2d

Construct a non-overlapping Conv2d from the given parameters.

Parameters:

Name	Type	Description	Default
color_channels	int	The number of color channels in the input image(s).	required
patch_size	int | tuple[int, int] | Sequence[int]	The dimension of the non-overlapping rectangular patches to tile the image with; if a single value is given, a square patch will be used.	required

Returns:

Type	Description
nn.Conv2d	The constructed Conv2d.