muti input cnn in matalb how to do that and how to feed the data in the model?

I want to train a muti input cnn in matalb how to do that and how to feed the data in the model?

like in python we do this:

history=final_model.fit(x=[X_insp_tr, X_exp_tr], y=y_tr,

epochs=100, batch_size=32,

validation_data=([X_insp_val, X_exp_val], y_val))

Answer 1

[Jaimin]

Hello @Arka Roy

You can utilise “connectLayers” function to build a cnn based deep learning model that accepts multiple inputs.

Kindly refer to the below code for sample model created using the “connectLayers” function.

% Define input layers for each input branch
inputLayer1 = imageInputLayer([28 28 1], 'Name', 'input1'); % Example size
inputLayer2 = imageInputLayer([28 28 1], 'Name', 'input2'); % Example size

% Define branches for each input with explicit layer names
branch1 = [
    inputLayer1
    convolution2dLayer(3, 8, 'Padding', 'same', 'Name', 'conv1_1')
    reluLayer('Name', 'relu1_1')
    maxPooling2dLayer(2, 'Stride', 2, 'Name', 'maxpool1_1')
];

branch2 = [
    inputLayer2
    convolution2dLayer(3, 8, 'Padding', 'same', 'Name', 'conv1_2')
    reluLayer('Name', 'relu1_2')
    maxPooling2dLayer(2, 'Stride', 2, 'Name', 'maxpool1_2')
];

% Concatenate branches
concatLayer = concatenationLayer(3, 2, 'Name', 'concat');

% Define the rest of the network
finalLayers = [
    fullyConnectedLayer(10, 'Name', 'fc')
    softmaxLayer('Name', 'softmax')
    classificationLayer('Name', 'classoutput')
];

% Assemble the network
layers = layerGraph(branch1);
layers = addLayers(layers, branch2);
layers = addLayers(layers, concatLayer);
layers = connectLayers(layers, 'maxpool1_1', 'concat/in1');
layers = connectLayers(layers, 'maxpool1_2', 'concat/in2');
layers = addLayers(layers, finalLayers);
layers = connectLayers(layers, 'concat', 'fc');

For more information on “connectLayers” function, kindly refer the following MathWorks Documentation:

https://www.mathworks.com/help/deeplearning/ref/nnet.cnn.layergraph.connectlayers.html

I hope this will be helpful.

Answer 2

[Arka Roy]

Thank you for helping me.

Just another query...if you please help on that...

Now, lets say i have traiing data={X_train1, X_train2, y_train} and validation data={X_val1, X_val2, y_val}. Then how to feed these data in trainnetwork to feed to model

In python i did this. How to do it in matlab?

history=final_model.fit(x=[X_insp_tr, X_exp_tr], y=y_tr,

epochs=100, batch_size=32,

validation_data=([X_insp_val, X_exp_val], y_val))

[Jaimin]

You can utilize "imageDatastore" and "arrayDatastore" to read images and labels. Once done, you can merge them using the "combine" function.

Kindly refer to the snippet below.

% Combine data into a datastore
dsTrain = combine(imageDatastore(X_insp_tr), imageDatastore(X_exp_tr), arrayDatastore(y_tr));

dsVal = combine(imageDatastore(X_insp_val), imageDatastore(X_exp_val), arrayDatastore(y_val));

For more information on "imageDatastore", "arrayDatastore" and "combine" function, kindly refer the following MathWorks Documentation:

imageDatastore: https://www.mathworks.com/help/matlab/ref/matlab.io.datastore.imagedatastore.html

arrayDatastore: https://www.mathworks.com/help/matlab/ref/matlab.io.datastore.arraydatastore.html

combine: https://www.mathworks.com/help/bioinfo/ref/bioread.combine.html

I hope this will be helpful.