Hi all, how to create image datasets. I need them to train neural networks. I have about 15 to 20 images and I need to turn these images into an image dataset. Please.

I have tried to find the way to build image dataset but all of the example are using Python. But i want to use Matlab. Please help me.

Answer 1

[Abhijit Bhattacharjee]

This is easy to do in MATLAB! You can put all your images into a folder and use the imageDatastore command.

Assuming the folder of images is on the path, here is an example:

imds = imageDatastore("name_of_image_folder");

[Nurul Farhana Mohd Fadzli]

After that what i need to do ? can you show me the example of your matlab interface?

[Abhijit Bhattacharjee]

What you do next depends on your application. In your original question, you asked what you need to make a dataset. The code I provided should be sufficient for that.

Answer 2

[yanqi liu]

yes，sir，may be use cnn transfer to train model，such as

unzip('MerchData.zip');
% use image folder to get dataset
imds = imageDatastore('MerchData','IncludeSubfolders',true,'LabelSource','foldernames');
[imdsTrain,imdsValidation] = splitEachLabel(imds,0.7,'randomized');
% use Alexnet to get cnn model
alex_net = alexnet;
class_number = length(unique(imds.Labels));
alex_net_share = alex_net.Layers(1:end-3);
alex_net_add = [
    fullyConnectedLayer(class_number,'Name','fc8','WeightLearnRateFactor',10, 'BiasLearnRateFactor',20)
    softmaxLayer('Name','softmax')
    classificationLayer('Name','classification')
    ];
layers_1 = [alex_net_share
    alex_net_add];
% train
augimdsTrain = augmentedImageDatastore([227 227],imdsTrain);
augimdsValidation = augmentedImageDatastore([227 227],imdsValidation);
miniBatchSize = 10;
valFrequency = floor(numel(augimdsTrain.Files)/miniBatchSize);
options = trainingOptions('sgdm', ...
    'MiniBatchSize',miniBatchSize, ...
    'MaxEpochs',5, ...
    'InitialLearnRate',3e-4, ...
    'Shuffle','every-epoch', ...
    'ValidationData',augimdsValidation, ...
    'ValidationFrequency',valFrequency, ...
    'Verbose',false);
trainedNet = trainNetwork(augimdsTrain,layers_1,options);
% test
[YPred,probs] = classify(trainedNet,augimdsValidation);
accuracy = mean(YPred == imdsValidation.Labels)

accuracy = 1

% app
idx = randperm(numel(imdsValidation.Files),4);
figure
for i = 1:4
    subplot(2,2,i)
    I = readimage(imdsValidation,idx(i));
    imshow(I)
    label = YPred(idx(i));
    title(string(label) + ", " + num2str(100*max(probs(idx(i),:)),3) + "%");
end

[Nurul Farhana Mohd Fadzli]

So, if i want to use my dataset, then what part do i need to change? Is it the MerchData? I am sorry im still new to Matlab.

[yanqi liu]

yes，sir，let us check the folder MerchData，we can find that one subfolder is one class，so if use our data，we can just make a new subfolder， and use name as subfolder name

then put images in it，and run code