deepboof.misc

Class TensorOps

java.lang.Object

deepboof.misc.TensorOps

public class TensorOps
extends java.lang.Object

Constructor Summary

Constructors

Constructor and Description
TensorOps()

Method Summary

Modifier and Type	Method and Description
static <T extends Tensor> T	AD(T input) Adds a dimension to the input tensor.
static void	boundSpatial(int[] bounds, int rows, int cols)
static void	checkShape(java.lang.String which, int tensor, int[] expected, int[] actual, boolean ignoreAxis0) Checks to see if the two tensors have the same shape, with the option to ignore the first axis for the 'actual' shape.
static void	checkShape(java.lang.String which, java.util.List<int[]> expected, java.util.List<Tensor<?>> actual, boolean ignoreAxis0) Compares a list of tensors shape's against each other.
static void	checkShape(Tensor_F32 a, Tensor_F32 b) Checks to see if the two tensors have the same shape
static void	checkShape(Tensor_F64 a, Tensor_F64 b) Checks to see if the two tensors have the same shape
static double	elementSum(Tensor tensor) Computes the sum of all the elements in the tensor
static void	fill(Tensor t, double value)
static int	outerLength(int[] shape, int startDimen) Computes the number of elements for an inner portion of the tensor starting at the specified index and going outside
static java.io.File	pathToRoot()
static int	sumTensorLength(java.util.List<int[]> shapes) Returns the total length of all the tensors in the list summed together
static int	tensorLength(int... shape) Returns the total length of one tensor
static int[]	TH(int[] elements) Truncates the head (element 0) from the array
static java.lang.String	toStringShape(int[] shape)
static java.lang.String	toStringShapeA(int[] shape)
static int[]	WI(int... elements) Convenience function for making it easy to create an array of ints
static int[]	WI(int[] elements, int a) Convenience function for making it easy to create an array of ints
static int[]	WI(int a, int[] elements) Convenience function for making it easy to create an array of ints
static java.util.List<int[]>	WI(int a, java.util.List<int[]> list)
static <T> java.util.List<T>	WT(T... elements) Convenience function for wrapping passed in elements into a list

Methods inherited from class java.lang.Object

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

Constructor Detail

TensorOps

public TensorOps()

Method Detail

WT

public static <T> java.util.List<T> WT(T... elements)

Convenience function for wrapping passed in elements into a list

WI

public static int[] WI(int... elements)

Convenience function for making it easy to create an array of ints

WI

public static int[] WI(int a,
                       int[] elements)

Convenience function for making it easy to create an array of ints

Returns:

[ a , elements[0] , ..., elements[N-1] ]

WI

public static int[] WI(int[] elements,
                       int a)

Convenience function for making it easy to create an array of ints

Returns:

[ elements[0] , ..., elements[N-1], a ]

TH

public static int[] TH(int[] elements)

Truncates the head (element 0) from the array

WI

public static java.util.List<int[]> WI(int a,
                                       java.util.List<int[]> list)

AD

public static <T extends Tensor> T AD(T input)

Adds a dimension to the input tensor. Returns a new tensor which references the same data as the original. Input Shape = [5 4 3] Output Shape = [1 5 4 3]

sumTensorLength

public static int sumTensorLength(java.util.List<int[]> shapes)

Returns the total length of all the tensors in the list summed together

Parameters:

shapes - List of tensor shapes

Returns:

Sum of tensor lengths

tensorLength

public static int tensorLength(int... shape)

Returns the total length of one tensor

Parameters:

shape - shape of a tensor

Returns:

length of the Tensor's data arrayn

checkShape

public static void checkShape(java.lang.String which,
                              java.util.List<int[]> expected,
                              java.util.List<Tensor<?>> actual,
                              boolean ignoreAxis0)

Compares a list of tensors shape's against each other. This simply invokes checkShape(String, int, int[], int[], boolean)

Parameters:

which - String describing which variables are being checked

expected - List of expected tensors

actual - List of actual tensors. Axis 0 is optionally ignored here, see ignoreAxis0

ignoreAxis0 - true to ignore axis 0

checkShape

public static void checkShape(Tensor_F64 a,
                              Tensor_F64 b)

Checks to see if the two tensors have the same shape

Parameters:

a - tensor

b - tensor

checkShape

public static void checkShape(Tensor_F32 a,
                              Tensor_F32 b)

Checks to see if the two tensors have the same shape

Parameters:

a - tensor

b - tensor

checkShape

public static void checkShape(java.lang.String which,
                              int tensor,
                              int[] expected,
                              int[] actual,
                              boolean ignoreAxis0)

Checks to see if the two tensors have the same shape, with the option to ignore the first axis for the 'actual' shape. The first axis is typically the mini-batch, but the expected value might not include the mini-batch since the size of the mini-batch is determined later on. Throws an IllegalArgumentException if they don't match.

Parameters:

which - String describing which variable is being checked

tensor - Index of the tensor in a tensor list. Used to provide a more detailed error message. If < 0 then this is ignored

expected - Expected shape.

actual - Actual shape. Axis 0 is optionally ignored.

ignoreAxis0 - If true it will ignore the first dimension in expected

toStringShapeA

public static java.lang.String toStringShapeA(int[] shape)

toStringShape

public static java.lang.String toStringShape(int[] shape)

outerLength

public static int outerLength(int[] shape,
                              int startDimen)

Computes the number of elements for an inner portion of the tensor starting at the specified index and going outside

Example:
Tensor shape = (d[0], ... , d[K-1]). Then if start dimen is 2, the output will be the product of d[2] to d[K-1].

pathToRoot

public static java.io.File pathToRoot()

elementSum

public static double elementSum(Tensor tensor)

Computes the sum of all the elements in the tensor

Parameters:

tensor - Tensor

fill

public static void fill(Tensor t,
                        double value)

boundSpatial

public static void boundSpatial(int[] bounds,
                                int rows,
                                int cols)