twodlearn.convnet module

class twodlearn.convnet.Conv1x1Proj(trainable=True, name=None, *args, **kwargs)

Bases: twodlearn.core.layers.Layer

Tdl autoinitialization with arguments:

_linop

(Submodel)

kernel

(ParameterInit) Autoinit with arguments [‘initializer’, ‘trainable’]

input_shape

(InputArgument)

activation

(InputArgument)

units

(InputArgument) Number of output units (int).

use_bias

(InputArgument)

bias

(ParameterInit) Autoinit with arguments [‘initializer’, ‘trainable’, ‘use_bias’]

activation

bias

Autoinit with arguments [‘initializer’, ‘trainable’, ‘use_bias’]

call(inputs)

This is where the layer’s logic lives.

Parameters

• inputs – Input tensor, or list/tuple of input tensors.

• **kwargs – Additional keyword arguments.

Returns

A tensor or list/tuple of tensors.

compute_output_shape(input_shape=None)

Computes the output shape of the layer.

Assumes that the layer will be built to match that input shape provided.

Parameters

input_shape – Shape tuple (tuple of integers) or list of shape tuples (one per output tensor of the layer). Shape tuples can include None for free dimensions, instead of an integer.

Returns

An input shape tuple.

get_transpose(use_bias=None, activation=None, trainable=True)

kernel

Autoinit with arguments [‘initializer’, ‘trainable’]

units

Number of output units (int).

use_bias

class twodlearn.convnet.Conv2DLayer(trainable=True, name=None, *args, **kwargs)

Bases: twodlearn.core.layers.Layer

Tdl autoinitialization with arguments:

padding

(InputArgument) Padding for the convolution. It could be either ‘valid’ or ‘same’. Default is ‘valid’.

kernel

(ParameterInit) Autoinit with arguments [‘initializer’, ‘trainable’]

input_shape

(InputArgument) Input tensor shape: (n_samples, n_rows, n_cols, n_channels).

kernel_size

(InputArgument) Size of the convolution kernels. Must be a tuple/list of two elements (height, width)

dilation_rate

(InputArgument) Defaults to (1, 1).

filters

(InputArgument) Number of filters (int), equal to the number of output maps.

use_bias

(InputArgument)

bias

(ParameterInit) Autoinit with arguments [‘initializer’, ‘trainable’, ‘use_bias’]

strides

(InputArgument) Convolution strides. Default is (1, 1).

bias

Autoinit with arguments [‘initializer’, ‘trainable’, ‘use_bias’]

call(inputs, *args, **kargs)

This is where the layer’s logic lives.

Parameters

• inputs – Input tensor, or list/tuple of input tensors.

• **kwargs – Additional keyword arguments.

Returns

A tensor or list/tuple of tensors.

compute_output_shape(input_shape=None)

Computes the output shape of the layer.

Assumes that the layer will be built to match that input shape provided.

Parameters

input_shape – Shape tuple (tuple of integers) or list of shape tuples (one per output tensor of the layer). Shape tuples can include None for free dimensions, instead of an integer.

Returns

An input shape tuple.

dilation_rate

Defaults to (1, 1).

filters

Number of filters (int), equal to the number of output maps.

input_shape

(n_samples, n_rows, n_cols, n_channels).

Type

Input tensor shape

kernel

Autoinit with arguments [‘initializer’, ‘trainable’]

kernel_size

Size of the convolution kernels. Must be a tuple/list of two elements (height, width)

padding

Padding for the convolution. It could be either ‘valid’ or ‘same’. Default is ‘valid’.

strides

Convolution strides. Default is (1, 1).

use_bias

class twodlearn.convnet.Conv2DTranspose(trainable=True, name=None, *args, **kwargs)

Bases: twodlearn.convnet.Conv2DLayer

Tdl autoinitialization with arguments:

padding

(InputArgument) Padding for the convolution. It could be either ‘valid’ or ‘same’. Default is ‘valid’.

kernel

(ParameterInit) Autoinit with arguments [‘initializer’, ‘trainable’]

output_padding

(InputArgument)

input_shape

(InputArgument) Input tensor shape: (n_samples, n_rows, n_cols, n_channels).

kernel_size

(InputArgument) Size of the convolution kernels. Must be a tuple/list of two elements (height, width)

dilation_rate

(InputArgument) Defaults to (1, 1).

filters

(InputArgument) Number of filters (int), equal to the number of output maps.

use_bias

(InputArgument)

bias

(ParameterInit) Autoinit with arguments [‘initializer’, ‘trainable’, ‘use_bias’]

strides

(InputArgument) Convolution strides. Default is (1, 1).

call(inputs, *args, **kargs)

This is where the layer’s logic lives.

Parameters

• inputs – Input tensor, or list/tuple of input tensors.

• **kwargs – Additional keyword arguments.

Returns

A tensor or list/tuple of tensors.

compute_output_shape(input_shape)

Computes the output shape of the layer.

Assumes that the layer will be built to match that input shape provided.

Parameters

input_shape – Shape tuple (tuple of integers) or list of shape tuples (one per output tensor of the layer). Shape tuples can include None for free dimensions, instead of an integer.

Returns

An input shape tuple.

kernel

Autoinit with arguments [‘initializer’, ‘trainable’]

output_padding

static transpose_output_lenght(input_length, filter_size, padding, output_padding=None, stride=0, dilation=1)

twodlearn.convnet.conv_output_length(input_length, filter_size, padding, stride, dilation=1)

Determines output length of a convolution given input length.

Parameters

• input_length (int) – integer.

• filter_size (int) – integer.

• padding (str) – one of “same”, “valid”, “full”, “causal”

• stride (int) – integer.

• dilation (int) – dilation rate, integer.

Returns

the output length.

Return type

int