Class ConvTranspose

Represents a ConvTranspose transpose convolution layer, which applies a convolution filter to an input tensor.

Inheritance
Object
Layer
FusedActivation
ConvTranspose
Inherited Members
FusedActivation.fusedActivation
Layer.inputs
Layer.name
Layer.outputs
Layer.flags
Namespace: Unity.Sentis.Layers
Syntax
[Serializable]
public class ConvTranspose : FusedActivation

Constructors

ConvTranspose(String, String, String, Int32[], Int32[], AutoPad, Int32[], Int32[], FusableActivation)

Declaration
public ConvTranspose(string name, string input, string kernel, int[] strides, int[] pads, AutoPad autoPad, int[] outputPadding, int[] kernelShape = null, FusableActivation fusedActivation = FusableActivation.None)
Parameters
Type Name Description
String name
String input
String kernel
Int32[] strides
Int32[] pads
AutoPad autoPad
Int32[] outputPadding
Int32[] kernelShape
FusableActivation fusedActivation

ConvTranspose(String, String, String, String, Int32[], Int32[], AutoPad, Int32[], Int32[], FusableActivation)

Initializes and returns an instance of ConvTranspose convolution layer.

Declaration
public ConvTranspose(string name, string input, string kernel, string bias, int[] strides, int[] pads, AutoPad autoPad, int[] outputPadding, int[] kernelShape = null, FusableActivation fusedActivation = FusableActivation.None)
Parameters
Type Name Description
String name

The name to use for the output tensor of the layer.
String input

The name to use for the input tensor of the layer.
String kernel

The name to use for the filter tensor of the layer.
String bias

The name to use for the optional bias tensor of the layer.
Int32[] strides

The optional stride value for each spatial dimension of the filter.
Int32[] pads

The optional lower and upper padding values for each spatial dimension of the filter.
AutoPad autoPad

The auto padding mode of the convolution.
Int32[] outputPadding

The output padding value for each spatial dimension in the filter.
Int32[] kernelShape
FusableActivation fusedActivation

The fused activation type to apply after the convolution. The default value is None.

Fields

autoPad

The auto padding mode of the transpose convolution.

Declaration
public AutoPad autoPad
Field Value
Type Description
AutoPad

kernelShape

Declaration
public int[] kernelShape
Field Value
Type Description
Int32[]

outputPadding

The output padding value for each spatial dimension in the filter.

The layer adds the output padding to the side with higher coordinate indices in the output tensor.

If this is null the layer uses a default of [0, 0, ..., 0].

Declaration
public int[] outputPadding
Field Value
Type Description
Int32[]

pads

The lower and upper padding values for each spatial dimension of the filter. For example [pad_left, pad_right] for 1D, or [pad_top, pad_bottom, pad_left, pad_right] for 2D.

If this is null the layer uses a default of [0, 0, ..., 0].

Declaration
public int[] pads
Field Value
Type Description
Int32[]

strides

The stride value for each spatial dimension of the filter.

If this is null the layer uses a default of [1, 1, ..., 1].

Declaration
public int[] strides
Field Value
Type Description
Int32[]

Methods

Execute(Tensor[], ExecutionContext)

Executes the layer using the operations and variables from the ExecutionContext and returns the output tensor.

If the layer has more than one output, output tensors are saved to variables.

Declaration
public override Tensor Execute(Tensor[] inputTensors, ExecutionContext ctx)
Parameters
Type Name Description
Tensor[] inputTensors
ExecutionContext ctx
Returns
Type Description
Tensor
Overrides

ToString()

Returns a string that represents the Layer.

Declaration
public override string ToString()
Returns
Type Description
String
Overrides