deepboof.impl.forward.standard

Class BaseSpatialPadding2D<T extends Tensor<T>>

java.lang.Object

deepboof.BaseTensor

deepboof.impl.forward.standard.BaseSpatialPadding2D<T>

All Implemented Interfaces:

SpatialPadding2D<T>, ITensor, VTensor

Direct Known Subclasses:

SpatialPadding2D_F32, SpatialPadding2D_F64

public abstract class BaseSpatialPadding2D<T extends Tensor<T>>
extends BaseTensor
implements SpatialPadding2D<T>

Abstract class fo all virtual 2D spatial padding implementation. Virtual padding contains a reference to the original input tensor which is going to be padded and on the fly will generate the values for elements which are not explicitly contained in the input tensor. This can reduce memory consumption and is more simplistic to implement for more complex padding methods.

Field Summary

Fields

Modifier and Type	Field and Description
protected int	COL0
protected int	COL1
protected ConfigPadding	config
protected T	input
protected int	ROW0
protected int	ROW1

Fields inherited from class deepboof.BaseTensor

shape

Constructor Summary

Constructors

Constructor and Description
BaseSpatialPadding2D(ConfigPadding config)

Method Summary

Modifier and Type	Method and Description
<T extends Tensor<T>> void	checkBackwardsShapeChannel(Tensor<T> padded, Tensor<T> original) Sanity checks the input for backwards images
<T extends Tensor<T>> void	checkBackwardsShapeImage(Tensor<T> padded, Tensor<T> original) Sanity checks the input for backwards images
int	getPaddingCol0() Returns the lower-extent padding along the tensor's columns.
int	getPaddingCol1() Returns the upper-extent padding along the tensor's columns.
int	getPaddingRow0() Returns the lower-extent padding along the tensor's rows.
int	getPaddingRow1() Returns the upper-extent padding along the tensor's rows.
void	setInput(T input) Spatial tensor that padding is being added around
int[]	shapeGivenInput(int... inputShape) Returns what the tensor's shape will be when given can input tensor with the spcified shape.

Methods inherited from class deepboof.BaseTensor

createCoor, getDimension, getShape, isShape, length

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Methods inherited from interface deepboof.forward.SpatialPadding2D

getClippingOffsetCol, getClippingOffsetRow, getTensorType, isClipped

Methods inherited from interface deepboof.ITensor

createCoor, getDataType, getDimension, getShape, isShape, length

Field Detail

config

protected ConfigPadding config

input

protected T extends Tensor<T> input

ROW0

protected int ROW0

ROW1

protected int ROW1

COL0

protected int COL0

COL1

protected int COL1

Constructor Detail

BaseSpatialPadding2D

public BaseSpatialPadding2D(ConfigPadding config)

Method Detail

setInput

public void setInput(T input)

Spatial tensor that padding is being added around

Specified by:

setInput in interface SpatialPadding2D<T extends Tensor<T>>

Parameters:

input - The input tensor

getPaddingRow0

public int getPaddingRow0()

Returns the lower-extent padding along the tensor's rows.

Specified by:

getPaddingRow0 in interface SpatialPadding2D<T extends Tensor<T>>

Returns:

padding

getPaddingCol0

public int getPaddingCol0()

Returns the lower-extent padding along the tensor's columns.

Specified by:

getPaddingCol0 in interface SpatialPadding2D<T extends Tensor<T>>

Returns:

padding

getPaddingRow1

public int getPaddingRow1()

Returns the upper-extent padding along the tensor's rows.

Specified by:

getPaddingRow1 in interface SpatialPadding2D<T extends Tensor<T>>

Returns:

padding

getPaddingCol1

public int getPaddingCol1()

Returns the upper-extent padding along the tensor's columns.

Specified by:

getPaddingCol1 in interface SpatialPadding2D<T extends Tensor<T>>

Returns:

padding

shapeGivenInput

public int[] shapeGivenInput(int... inputShape)

Returns what the tensor's shape will be when given can input tensor with the spcified shape.

Specified by:

shapeGivenInput in interface SpatialPadding2D<T extends Tensor<T>>

Parameters:

inputShape - Input spatial tensor. 3-DOF with no mini-batch or 4-DOF with mini-batch

Returns:

Tensor's shape

checkBackwardsShapeChannel

public <T extends Tensor<T>> void checkBackwardsShapeChannel(Tensor<T> padded,
                                                             Tensor<T> original)

Sanity checks the input for backwards images

Parameters:

padded - Input padded single channel image from forward layer

original - Output backwards spatial tensor

checkBackwardsShapeImage

public <T extends Tensor<T>> void checkBackwardsShapeImage(Tensor<T> padded,
                                                           Tensor<T> original)

Sanity checks the input for backwards images

Parameters:

padded - Input padded image from forward layer

original - Output backwards spatial tensor