how to use 'createIma​gePatchesF​romHypercu​be' function

2021 6 28
2 个回答
回答已采纳
更新时间：2022 7 8
4 次查看（30 天）

0 个投票

İ was trying to use 'createImagePatchesFromHypercube' function but there is an error like this;
'createImagePatchesFromHypercube' is used in Classify Hyperspectral Images Using Deep Learning.
İ reinstall 4 times library but i cant solve.
How i can figure out this.
Thanks for help

请先登录，再回答此问题。

 采纳的回答

Subhadeep Koley
Subhadeep Koley 2021-7-1
编辑：Subhadeep Koley 2021-7-1

1 个投票

It seems you're trying to execute createImagePatchesFromHypercube outside of the "Classify Hyperspectral Images Using Deep Learning" example. createImagePatchesFromHypercube is a helper function and preferably should not be called outside the specified example.
But, if you want to use it, you can navigate to the folder path, where the function resides. You can directly reach to that path by executing the following command,
openExample('deeplearning_shared/HyperspectralImageClassificationUsingDeepLearningExample')
Hope this helps.

更多回答（1 个）

0 个投票

Hey good morning guys,
I solved the problem, what I did was to copy the function code at the end of the code that we want to run, apparently matlab can't found the function file...
Here is the code:
% Example of HSI image segmentation
% Load the indian_pines data set consisting of a single hyperspectral
% image of size 145X145 with 220 color channels
hcube = hypercube("indian_pines.dat");
rgbImg = colorize(hcube,method="rgb");
imshow(rgbImg)
% Load the ground truth labels and specify the number of classes
gtLabel = load("indian_pines_gt.mat");
gtLabel = gtLabel.indian_pines_gt;
numClasses = 16;
% Preparing the train data, this reduces the number of spectral brands to
% the 30 most representative
dimReduction = 30;
imageData = hyperpca(hcube,dimReduction);
% Normalize the data
sd = std(imageData,[],3);
imageData = imageData./sd;
% Split the hyperspectral image into patches of size 25-by-25 pixels with
% 30 channels, also returns a single label for each patch, which is the label of the central pixel.
windowSize = 25;
inputSize = [windowSize windowSize dimReduction];
[allPatches,allLabels] = createImagePatchesFromHypercube(imageData,gtLabel,windowSize);
indianPineDataTransposed = permute(allPatches,[2 3 4 1]);
dsAllPatches = augmentedImageDatastore(inputSize,indianPineDataTransposed,allLabels);
% Not all of the cubes in this data set have labels. However, training the network
% requires labeled data. Select only the labeled cubes for training. Count how many labeled
% patches are available.
patchesLabeled = allPatches(allLabels>0,:,:,:);
patchLabels = allLabels(allLabels>0);
numCubes = size(patchesLabeled,1);
% Convert the numerical labels to categorical
patchLabels = categorical(patchLabels);
% Divide the patches into training and test data sets
[trainingIdx,valIdx,testIdx] = dividerand(numCubes,0.3,0,0.7);
dataInputTrain = patchesLabeled(trainingIdx,:,:,:);
dataLabelTrain = patchLabels(trainingIdx,1);
dataInputTest = patchesLabeled(testIdx,:,:,:);
dataLabelTest = patchLabels(testIdx,1);
% Transpose the input data.
dataInputTransposeTrain = permute(dataInputTrain,[2 3 4 1]);
dataInputTransposeTest = permute(dataInputTest,[2 3 4 1]);
% Create datastores that read batches of training and test data.
dsTrain = augmentedImageDatastore(inputSize,dataInputTransposeTrain,dataLabelTrain);
dsTest = augmentedImageDatastore(inputSize,dataInputTransposeTest,dataLabelTest);
%function that creates patches in the image
function [patchData,patchLabel] = createImagePatchesFromHypercube(hcube,groundTruthLabel,winSize)
padding = floor((winSize-1)/2);
zeroPaddingPatch = padarray(hcube,[padding,padding],0,'both');
[rows,cols,ch] = size(hcube);
patchData = zeros(rows*cols,winSize,winSize,ch);
patchLabel = zeros(rows*cols,1);
zeroPaddedInput = size(zeroPaddingPatch);
patchIdx = 1;
for i= (padding+1):(zeroPaddedInput(1)-padding)
for j= (padding+1):(zeroPaddedInput(2)-padding)
patch = zeroPaddingPatch(i-padding:i+padding,j-padding:j+padding,:);
patchData(patchIdx,:,:,:) = patch;
patchLabel(patchIdx,1) = groundTruthLabel(i-padding,j-padding);
patchIdx = patchIdx+1;
end
end
end

类别

在 帮助中心 和 File Exchange 中查找有关 Hyperspectral Image Processing 的更多信息

产品

标签

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by