dlresize

R2026b

Resize spatial dimensions of dlarray object

Syntax

Y = dlresize(X,'Scale',scale)

Y = dlresize(X,'OutputSize',outputSize)

Y = dlresize(___,Name=Value)

Description

Y = dlresize(X,'Scale',scale) resizes the spatial dimensions of the dlarray object X by a scale factor scale.

This function requires Deep Learning Toolbox™.

Y = dlresize(X,'OutputSize',outputSize) resizes the spatial dimensions of the dlarray object X so that the spatial dimension sizes are equal to outputSize.

Y = dlresize(___,Name=Value) adjusts the resizing operation using name-value arguments. If X is not a formatted dlarray, then you must specify the DataFormat name-value argument.

Examples

Resize dlarray by Scale Factor

Read an RGB image.

A = imread("peppers.png");

Convert the image to data type single for use in a dlarray. Then, create a dlarray containing the input image.

A = im2single(A);
dlarrayA = dlarray(A,"SSC");

Rescale the dlarray by a factor of 1.5 vertically.

dlarrayB = dlresize(dlarrayA,Scale=[1.5 1]);

Extract the image data from the resized dlarray B by using the extractdata (Deep Learning Toolbox) function.

B = extractdata(dlarrayB);

Display the original and resized images as a montage.

montage({A,B},ThumbnailSize=size(B,[1 2]), ...
    BorderSize=10,BackgroundColor="white")

Input Arguments

X — Deep learning array to resize

dlarray object

Deep learning array to resize, specified as a dlarray (Deep Learning Toolbox) object.

scale — Scale factor to resize input

positive number | vector of positive numbers

Scale factor to resize input, specified as a positive number or a vector of positive numbers of length equal to the number of spatial dimensions in X. If scale is a scalar, then dlresize applies the same scale factor to all spatial dimensions.

outputSize — Output size of resized input

vector of positive integers

Output size of resized input, specified as a vector of positive integers of length equal to the number of spatial dimensions in X. You can specify one element as a positive integer and specify the other elements as NaN, in which case the layer computes the other elements automatically to preserve the aspect ratio of the input.

Name-Value Arguments

Specify optional pairs of arguments as Name1=Value1,...,NameN=ValueN, where Name is the argument name and Value is the corresponding value. Name-value arguments must appear after other arguments, but the order of the pairs does not matter.

Example: Y = dlresize(X, Scale=[2 2], Method="linear") resizes the spatial dimensions of X by a factor of 2 in each spatial direction using linear interpolation

DataFormat — Dimension labels

string scalar | character vector

Dimension labels of the input dlarray object X, specified as a string scalar or character vector. Each character must be one of these labels:

• S — Spatial

• C — Channel

• B — Batch observations

• T — Time or sequence

• U — Unspecified

If X is not a formatted dlarray, then you must specify the DataFormat name-value argument. For more information, see dlarray (Deep Learning Toolbox).

Example: 'SSC' indicates the array has two spatial dimensions and one channel dimension, appropriate for 2-D RGB image data.

Method — Interpolation method

"nearest" (default) | "linear"

Interpolation method, specified as "nearest" for nearest neighbor interpolation or "linear" for bilinear interpolation.

GeometricTransformMode — Geometric transformation mode

"half-pixel" (default) | "asymmetric" | "aligned-corners"

Geometric transformation mode to map points from input space to output space, specified as "half-pixel", "asymmetric", or "aligned-corners". You can specify GeometricTransformMode as "aligned-corners" only if you specify Method as "linear".

NearestRoundingMode — Rounding mode for nearest neighbor interpolation

"round" (default) | "floor" | "onnx-10"

Rounding mode for nearest neighbor interpolation, specified as one of the following.

• "round" — use the same rounding behavior as the MATLAB^® round function.

• "floor" — use the same rounding behavior as the MATLAB floor function.

• "onnx-10" — reproduce the resizing behavior of the ONNX™ (Open Neural Network Exchange) opset 10 Resize operator.

This argument is used when you specify the Method as 'nearest'.

Output Arguments

Y — Resized deep learning array

dlarray object

Resized deep learning array, returned as a dlarray (Deep Learning Toolbox) object.

Extended Capabilities

C/C++ Code Generation

Generate C and C++ code using MATLAB® Coder™.

Usage notes and limitations:

• dlarray inputs with more than 3 spatial (S) dimensions are not supported.

• Code generation supports variable-size data for only the spatial (S) dimension of the dlarray input.

• For a dlarray input with more than one spatial dimension, all spatial dimensions must be fixed-size or all variable-size.

• DataFormat must be compile-time constant.

GPU Code Generation

Generate CUDA® code for NVIDIA® GPUs using GPU Coder™.

Usage notes and limitations:

• dlarray inputs with more than 3 spatial (S) dimensions are not supported.

• Code generation supports variable-size data for only the spatial (S) dimension of the dlarray input.

• For a dlarray input with more than one spatial dimension, all spatial dimensions must be fixed-size or all variable-size.

• DataFormat must be compile-time constant.

GPU Arrays

Accelerate code by running on a graphics processing unit (GPU) using Parallel Computing Toolbox™.

This function fully supports GPU arrays. For more information, see Image Processing on a GPU.

Version History

Introduced in R2020b

R2026a: Specify GeometricTransformMode as "aligned-corners"

Map points from input space to output space by aligning corners by specifying the GeometricTransformMode as "aligned-corners" when the interpolation method Method is set to "linear".

R2024a: Generate CUDA code using GPU Coder

dlresize now supports the generation of optimized CUDA^® code (requires GPU Coder™).

R2024a: Generate C code using MATLAB Coder

dlresize now supports the generation of C code (requires MATLAB Coder™).