deepboof.misc

Class TensorOps_F32

java.lang.Object

deepboof.misc.TensorOps_F32

public class TensorOps_F32
extends java.lang.Object

Constructor Summary

Constructors

Constructor and Description
TensorOps_F32()

Method Summary

Modifier and Type	Method and Description
static void	elementAdd(Tensor_F32 A, Tensor_F32 B, Tensor_F32 output) Performs element-wise addition between the two tensors and stores results in output.
static void	elementMult(Tensor_F32 tensor, float value) Performs an element-wise scalar multiplication on the tensor
static void	elementMult(Tensor_F32 input, float value, Tensor_F32 output) Performs an element-wise scalar multiplication
static void	elementMult(Tensor_F32 A, Tensor_F32 B, Tensor_F32 output) Performs element-wise multiplication between the two tensors and stores results in output.
static float	elementSum(Tensor_F32 tensor) Computes the sum of all the elements in the tensor
static void	fill(Tensor_F32 tensor, float value) Fills the tensor with the specified value
static void	fillSpatialBorder(Tensor_F32 tensor, int[] coor, int borderY0, int borderX0, int borderY1, int borderX1, float value) Fills the border with the specified value.
static void	insertSpatial(Tensor_F32 src, int[] srcCoor, Tensor_F32 dst, int[] dstCoor) Inserts the spatial region of one tensor into another.
static void	insertSubChannel(Tensor_F32 src, int srcStartIndex, int srcStride, Tensor_F32 dst, int dstStartIndex, int dstStride, int rows, int columns) Used to copy a sub-image between two image tensors.
static void	printSpatial(Tensor_F32 tensor, int batch, int channel) Prints a single batch and channel in a spatial tensor

Methods inherited from class java.lang.Object

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

Constructor Detail

TensorOps_F32

public TensorOps_F32()

Method Detail

elementMult

public static void elementMult(Tensor_F32 tensor,
                               float value)

Performs an element-wise scalar multiplication on the tensor

Parameters:

tensor - Tensor which is multiplied

value - value of the multiplication

elementMult

public static void elementMult(Tensor_F32 input,
                               float value,
                               Tensor_F32 output)

Performs an element-wise scalar multiplication

Parameters:

input - Tensor which is multiplied (not modified)

value - value of the multiplication (modified)

output - Tensor where the results are stored

elementMult

public static void elementMult(Tensor_F32 A,
                               Tensor_F32 B,
                               Tensor_F32 output)

Performs element-wise multiplication between the two tensors and stores results in output. All tensors must have the same shape.

output[i] = A[i]*B[i]

Parameters:

A - Input tensor. Can be the same as output.

B - Input tensor. Can be the same as output.

output - Output tensor.

elementAdd

public static void elementAdd(Tensor_F32 A,
                              Tensor_F32 B,
                              Tensor_F32 output)

Performs element-wise addition between the two tensors and stores results in output. All tensors must have the same shape.

output[i] = A[i] + B[i]

Parameters:

A - Input tensor. Can be the same as output.

B - Input tensor. Can be the same as output.

output - Output tensor.

elementSum

public static float elementSum(Tensor_F32 tensor)

Computes the sum of all the elements in the tensor

Parameters:

tensor - Tensor

insertSubChannel

public static void insertSubChannel(Tensor_F32 src,
                                    int srcStartIndex,
                                    int srcStride,
                                    Tensor_F32 dst,
                                    int dstStartIndex,
                                    int dstStride,
                                    int rows,
                                    int columns)

Used to copy a sub-image between two image tensors.

Parameters:

src - Source tensor

srcStartIndex - Start index in input tensor.

srcStride - Row-stride for input tensor

dst - Destination tensor

dstStartIndex - Start index in destination tensor.

dstStride - Row-stride for destination tensor

rows - Number of rows to be copied

columns - Number of columns to be copied

insertSpatial

public static void insertSpatial(Tensor_F32 src,
                                 int[] srcCoor,
                                 Tensor_F32 dst,
                                 int[] dstCoor)

Inserts the spatial region of one tensor into another. Both tensors are assumed to follow the following pattern for their shape. (..., C, H, W). C is for the number of channels, H is for the image's height, and W, is for the image's width.

Parameters:

src - Source tensor. Entire image is copied into dst.

srcCoor - Coordinate of spatial region. ( ..., 0, 0, 0) modified.

dst - Destination tensor. The source image can be smaller than the destination, but not larger.

dstCoor - Coordinate of spatial region. ( ..., 0, y, x) modified.

fillSpatialBorder

public static void fillSpatialBorder(Tensor_F32 tensor,
                                     int[] coor,
                                     int borderY0,
                                     int borderX0,
                                     int borderY1,
                                     int borderX1,
                                     float value)

Fills the border with the specified value. The tensor is assumed to have the following shape ( ... , C , H, W), C is for he number of channels, H is for the image's height, and W, is for the image's width.

Parameters:

tensor - Tensor with a spatial region at the end. Modified.

coor - Coordinate of the targeted spatial region inside the tensor. axis values for C,H,W are ignored. Modified.

borderY0 - Lower extent's border length along Y axis

borderX0 - Lower extent's border length along X axis

borderY1 - Upper extent's border length along Y axis

borderX1 - Upper extent's border length along X axis

value - Value that is to be inserted

printSpatial

public static void printSpatial(Tensor_F32 tensor,
                                int batch,
                                int channel)

Prints a single batch and channel in a spatial tensor

Parameters:

tensor - The tensor

batch - Batch number

channel - channel

fill

public static void fill(Tensor_F32 tensor,
                        float value)

Fills the tensor with the specified value

Parameters:

tensor - The tensor

value - fill value