Convolutional Layers#

Classes#

class Conv(net_out, shapes, *, filters=None, layer=None, factor=None, groups=1, kernel=3, stride=1, padding=0, dropout=0, activation='ELU', norm=None, padding_mode=None, **kwargs)[source]#

Bases: BaseSingleLayer

Convolutional layer constructor.

Supports 1D, 2D, and 3D convolution.

Attributes:

groupint: Layer group, if 0 it will always be used, else it will only be used if its group matches the Networks
descriptionstr: Description of the layer
layersSequential: Layers to loop through in the forward pass

Parameters:

net_outlist[int]: Shape of the network’s output, required only if layer contains factor
shapesShapes: Shape of the outputs from each layer
filtersint | NoneType, Optional: Number of convolutional filters, will be used if provided, else factor will be used
layerint | NoneType, Optional: If factor is not None, which layer for factor to be relative to, if None, network output will be used
factorfloat | NoneType, Optional: Number of convolutional filters equal to the output channels, or if layer is provided, the layer’s channels, multiplied by factor, won’t be used if filters is provided
groupsint, Optional: Number of input channel groups, each with its own convolutional filter(s), input and output channels must both be divisible by the number of groups, default = 1
kernelint | list[int], Optional: Size of the kernel, default = 3
strideint | list[int], Optional: Stride of the kernel, default = 1
paddingint | {‘same’} | list[int], Optional: Input padding, can an int, list of ints or ‘same’ where ‘same’ preserves the input shape, default = 0
dropoutfloat, Optional: Probability of dropout, default = 0
activationstr | NoneType, Optional: Which activation function to use from PyTorch, default = ‘ELU’
norm{NoneType, ‘batch’, ‘layer’}: If batch or layer normalisation should be used
padding_mode{NoneType, ‘zeros’, ‘reflect’, ‘replicate’, ‘circular’}: Padding mode to use from PyTorch, if None zero padding is used if version is >3.9.4 else replication padding is used
**kwargs: Leftover parameters to pass to base layer for checking

class ConvDepth(net_out, shapes, *, filters=None, layer=None, factor=None, kernel=3, stride=1, padding=0, dropout=0, activation='ELU', norm=None, padding_mode=None, **kwargs)[source]#

Bases: Conv

Constructs a depthwise convolutional layer.

Supports 1D, 2D, and 3D convolution.

Attributes:

groupint: Layer group, if 0 it will always be used, else it will only be used if its group matches the Networks
descriptionstr: Description of the layer
layersSequential: Layers to loop through in the forward pass

Parameters:

net_outlist[int]: Shape of the network’s output, required only if layer contains factor
shapesShapes: Shape of the outputs from each layer
filtersint | NoneType, Optional: Number of convolutional filters, will be used if provided, else factor will be used
layerint | NoneType, Optional: If factor is not None, which layer for factor to be relative to, if None, network output will be used
factorfloat | NoneType, Optional: Number of convolutional filters equal to the output channels multiplied by factor, won’t be used if filters is provided
kernelint | list[int], Optional: Size of the kernel, default = 3
strideint | list[int], Optional: Stride of the kernel, default = 1
paddingint | {‘same’} | list[int], Optional: Input padding, can an int, list of ints or ‘same’ where ‘same’ preserves the input shape, default = 0
dropoutfloat, Optional: Probability of dropout, default = 0
activationstr | NoneType, Optional: Which activation function to use, default = ‘ELU’
norm{NoneType, ‘batch’, ‘layer’}: If batch or layer normalisation should be used
padding_mode{NoneType, ‘zeros’, ‘reflect’, ‘replicate’, ‘circular’}: Padding mode to use from PyTorch, if None zero padding is used if version is >3.9.4 else replication padding is used
**kwargs: Leftover parameters to pass to base layer for checking

class ConvDepthDownscale(net_out, shapes, *, dropout=0, activation='ELU', norm=None, padding_mode=None, **kwargs)[source]#

Bases: Conv

Constructs depth downscaler using convolution with kernel size of 1.

Attributes:

groupint: Layer group, if 0 it will always be used, else it will only be used if its group matches the Networks
descriptionstr: Description of the layer
layersSequential: Layers to loop through in the forward pass

Parameters:

net_outlist[int]: Shape of the network’s output, required only if layer contains factor
shapes: Shapes: Shape of the outputs from each layer
dropoutfloat, Optional: Probability of dropout, default = 0
activationstr | NoneType, Optional: Which activation function to use, default = ‘ELU’
norm{NoneType, ‘batch’, ‘layer’}: If batch or layer normalisation should be used
padding_mode{NoneType, ‘zeros’, ‘reflect’, ‘replicate’, ‘circular’}: Padding mode to use from PyTorch, if None zero padding is used if version is >3.9.4 else replication padding is used
**kwargs: Leftover parameters to pass to base layer for checking

class ConvDownscale(net_out, shapes, *, filters=None, layer=None, factor=None, scale=2, dropout=0, activation='ELU', norm=None, **kwargs)[source]#

Bases: Conv

Constructs a strided convolutional layer for downscaling.

The scale factor is equal to the stride and kernel size.

Attributes:

groupint: Layer group, if 0 it will always be used, else it will only be used if its group matches the Networks
descriptionstr: Description of the layer
layersSequential: Layers to loop through in the forward pass

Parameters:

net_outlist[int]: Shape of the network’s output, required only if layer contains factor
shapes: Shapes: Shape of the outputs from each layer
filtersint | NoneType, Optional: Number of convolutional filters, will be used if provided, else factor will be used
layerint | NoneType, Optional: If factor is not None, which layer for factor to be relative to, if None, network output will be used
factorfloat | NoneType, Optional: Number of convolutional filters equal to the output channels multiplied by factor, won’t be used if filters is provided
scaleint, Optional: Stride and size of the kernel, which acts as the downscaling factor, default = 2
dropoutfloat, Optional: Probability of dropout, default = 0
activationstr | NoneType, Optional: Which activation function to use, default = ‘ELU’
norm{NoneType, ‘batch’, ‘layer’}: If batch or layer normalisation should be used
**kwargs: Leftover parameters to pass to base layer for checking

class ConvTranspose(net_out, shapes, *, filters=None, layer=None, factor=None, kernel=3, stride=1, out_padding=0, dilation=1, padding=0, dropout=0, padding_mode='zeros', activation='ELU', norm=None, **kwargs)[source]#

Bases: BaseSingleLayer

Constructs a transpose convolutional layer with fractional stride for input upscaling.

Supports 1D, 2D, and 3D transposed convolution.

Attributes:

groupint: Layer group, if 0 it will always be used, else it will only be used if its group matches the Networks
descriptionstr: Description of the layer
layersSequential: Layers to loop through in the forward pass

Parameters:

net_outlist[int]: Shape of the network’s output, required only if layer contains factor
shapes: Shapes: Shape of the outputs from each layer
filtersint | NoneType, Optional: Number of convolutional filters, will be used if provided, else factor will be used
layerint | NoneType, Optional: If factor is not None, which layer for factor to be relative to, if None, network output will be used
factorfloat | NoneType, Optional: Number of convolutional filters equal to the output channels multiplied by factor, won’t be used if filters is provided
kernelint | list[int], Optional: Size of the kernel, default = 3
strideint | list[int], Optional: Stride of the kernel, default = 1
out_paddingint | list[int], Optional: Padding applied to the output, default = 0
dilationint | list[int], Optional: Spacing between kernel points, default = 1
paddingint | {‘same’} | list[int], Optional: Inverse of convolutional padding which removes rows from each dimension in the output, default = 0
dropoutfloat, Optional: Probability of dropout, default = 0
padding_mode{‘zeros’, ‘reflect’, ‘replicate’, ‘circular’}: Padding mode to use from PyTorch
activationstr | NoneType, Optional: Which activation function to use, default = ‘ELU’
norm{NoneType, ‘batch’, ‘layer’}: If batch or layer normalisation should be used
**kwargs: Leftover parameters to pass to base layer for checking

forward(x, *_, **__)[source]#

Forward pass of the transposed convolutional layer

Parameters:

xTensor: Input tensor with shape (N,…) and type float, where N is the batch size

Returns:

Tensor: Output tensor with shape (N,…) and type float

class ConvTransposeUpscale(net_out, shapes, *, filters=None, layer=None, factor=None, scale=2, out_padding=0, dropout=0, activation='ELU', norm=None, **kwargs)[source]#

Bases: ConvTranspose

Constructs an upscaler using a transposed convolutional layer.

Supports 1D, 2D, and 3D transposed convolutional upscaling.

Attributes:

groupint: Layer group, if 0 it will always be used, else it will only be used if its group matches the Networks
descriptionstr: Description of the layer
layersSequential: Layers to loop through in the forward pass

Parameters:

net_outlist[int]: Shape of the network’s output, required only if layer contains factor
shapes: Shapes: Shape of the outputs from each layer
filtersint | NoneType, Optional: Number of convolutional filters, will be used if provided, else factor will be used
layerint | NoneType, Optional: If factor is not None, which layer for factor to be relative to, if None, network output will be used
factorfloat | NoneType, Optional: Number of convolutional filters equal to the output channels multiplied by factor, won’t be used if filters is provided
scaleint | list[int], Optional: Stride and size of the kernel, which acts as the upscaling factor, default = 2
out_paddingint | list[int], Optional: Padding applied to the output, default = 0
dropoutfloat, Optional: Probability of dropout, default = 0
activationstr | NoneType, Optional: Which activation function to use, default = ‘ELU’
norm{NoneType, ‘batch’, ‘layer’}: If batch or layer normalisation should be used
**kwargs: Leftover parameters to pass to base layer for checking

class ConvUpscale(net_out, shapes, *, filters=None, layer=None, factor=None, scale=2, kernel=3, dropout=0, activation='ELU', norm=None, padding_mode=None, **kwargs)[source]#

Bases: Conv

Constructs an upscaler using a convolutional layer and pixel shuffling.

Supports 1D, 2D, and 3D convolutional upscaling.

See Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network by Shi et al. (2016) for details.

Attributes:

groupint: Layer group, if 0 it will always be used, else it will only be used if its group matches the Networks
descriptionstr: Description of the layer
layersSequential: Layers to loop through in the forward pass

Parameters:

net_outlist[int]: Shape of the network’s output, required only if layer contains factor
shapes: Shapes: Shape of the outputs from each layer
filtersint | NoneType, Optional: Number of convolutional filters, will be used if provided, else factor will be used
layerint | NoneType, Optional: If factor is not None, which layer for factor to be relative to, if None, network output will be used
factorfloat | NoneType, Optional: Number of convolutional filters equal to the output channels multiplied by factor, won’t be used if filters is provided
scaleint, Optional: Factor to upscale the input by, default = 2
kernelint | list[int], Optional: Size of the kernel, default = 3
dropoutfloat, Optional: Probability of dropout, default = 0
activationstr | NoneType, Optional: Which activation function to use, default = ‘ELU’
norm{NoneType, ‘batch’, ‘layer’}: If batch or layer normalisation should be used
padding_mode{NoneType, ‘zeros’, ‘reflect’, ‘replicate’, ‘circular’}: Padding mode to use from PyTorch, if None zero padding is used if version is >3.9.4 else replicate padding is used
**kwargs: Leftover parameters to pass to base layer for checking

class PixelShuffle(scale, *, shapes=None, **kwargs)[source]#

Bases: BaseLayer

Used for upscaling by scale factor \(r\) for an input \((N,C\times r^n,D_1,...,D_n)\) to an output \((N,C,D_1\times r,...,D_n\times r)\).

Equivalent to torch.nn.PixelShuffle, but for nD.

Attributes:

groupint: Layer group, if 0 it will always be used, else it will only be used if its group matches the Networks
descriptionstr: Description of the layer

Parameters:

scaleint: Upscaling factor
shapes: Shapes | None, Optional: Shape of the outputs from each layer
**kwargs: Leftover parameters to pass to base layer for checking

extra_repr()[source]#

Displays layer parameters when printing the network

Returns:

str: Layer parameters

forward(x, *_, **__)[source]#

Forward pass of pixel shuffle

Parameters:

xTensor: Input tensor with shape \((N,C\times r^n,D_1,...,D_n)\) and type float, where N is the batch size, C is the number of channels, r is the upscaling factor and \(D_n\) is the length of dimension n

Returns:

Tensor: Output tensor with shape \((N,C,D_1\times r,...,D_n\times r)\) and type float