dlpy.layers.Conv2DTranspose

class dlpy.layers.Conv2DTranspose(n_filters, height=None, width=None, output_size=None, name=None, act='AUTO', dropout=0, fcmp_act=None, include_bias=True, init='XAVIER', init_bias=0, mean=0, std=1, output_padding=None, output_padding_height=None, output_padding_width=None, padding=None, padding_height=None, padding_width=None, stride=None, stride_horizontal=None, stride_vertical=None, truncation_factor=None, src_layers=None, **kwargs)

Bases: dlpy.layers.Conv2d

Transpose Convolution layer in 2D

Parameters:
n_filters : int

Specifies the number of filters for the layer.

width : int

Specifies the width of the kernel.

height : int

Specifies the height of the kernel.

output_size : tuple

Specifies the shape of output. 3D tensor with shape: (n_filters, new_rows, new_cols)

stride : int, optional

Specifies the step size for the moving window of the kernel over the input data.

name : string, optional

Specifies the name of the convolution layer.

stride_horizontal : int, optional

Specifies the horizontal stride.

stride_vertical : int, optional

Specifies the vertical stride.

padding : int, optional

Specifies the padding size, assuming equal padding vertically and horizontally.

padding_width : int, optional

Specifies the length of the horizontal padding.

padding_height : int, optional

Specifies the length of the vertical padding.

act : string, optional

Specifies the activation function.
Valid Values: AUTO, IDENTITY, LOGISTIC, SIGMOID, TANH, RECTIFIER, RELU, SOFPLUS, ELU, LEAKY, FCMP
Default: AUTO

fcmp_act : string, optional

Specifies the FCMP activation function for the layer.

init : string, optional

Specifies the initialization scheme for the layer.
Valid Values: XAVIER, UNIFORM, NORMAL, CAUCHY, XAVIER1, XAVIER2, MSRA, MSRA1, MSRA2
Default: XAVIER

mean : float, optional

Specifies the mean value when the init parameter is set to NORMAL.

std : float, optional

Specifies the standard deviation value when the init parameter is set to NORMAL.

output_padding : int

Specifies the number of pixels to add to the right and bottom sides of the input to adjust the output shape. This parameter is used to resolve ambiguity that might be introduced when stride is larger than 1. If the input shape is 32x32, kernel size is 3x3, and stride is 2 on both sides, and padding is 1 on both sides, then the output shape is 63x63. Setting the output_paddings to 1 on both sides changes the output shape to 64x64.

output_padding_height : int

Specifies the number of pixels to be added to the bottom side of the input to adjust the output shape. This parameter is used to resolve ambiguity that might be introduced when the stride is larger than 1. If the input shape is 32x32, kernel size is 3x3, stride is 2 on both sides, and padding is 1 on both sides, then the output shape is 63x63. Setting output_paddings to 1 on both sides changes the output shape to 64x64.

output_padding_width : int

specifies the number of pixels to add to the right side of the input to adjust the output shape. This parameter is used to resolve ambiguity that might be introduced when stride is larger than 1. If the input shape is 32x32, kernel size is 3x3, stride is 2 on both sides, and padding is 1 on both sides, then the output shape is 63x63. Setting the outputPaddings to 1 on both sides changes the output shape to 64x64.

truncation_factor : float, optional

Specifies the truncation threshold (truncationFactor x std), when the init parameter is set to NORMAL

init_bias : float, optional

Specifies the initial bias for the layer.

dropout : float, optional

Specifies the dropout rate.
Default: 0

include_bias : bool, optional

Includes bias neurons (default).

src_layers : iter-of-Layers, optional

Specifies the layers directed to this layer.

Returns:
Conv2DTranspose
__init__(n_filters, height=None, width=None, output_size=None, name=None, act='AUTO', dropout=0, fcmp_act=None, include_bias=True, init='XAVIER', init_bias=0, mean=0, std=1, output_padding=None, output_padding_height=None, output_padding_width=None, padding=None, padding_height=None, padding_width=None, stride=None, stride_horizontal=None, stride_vertical=None, truncation_factor=None, src_layers=None, **kwargs)

Initialize self. See help(type(self)) for accurate signature.

Methods

__init__(n_filters[, height, width, …]) Initialize self.
calculate_output_padding() calculate output_padding before adding the layer
count_instances()
format_name([block_num, local_count]) Format the name of the layer
get_number_of_instances()
to_model_params() Convert the model configuration to CAS action parameters