Untitled

% Load images and create an image datastore
faceData = imageDatastore({'Face1', 'Face2', 'Face3'}, 'IncludeSubfolders', true, 'LabelSource', 'foldernames');

% Define a function to resize images
resizeFcn = @(img) imresize(img, [224 224]);

% Apply the resize function to the datastore
faceData.ReadFcn = @(loc) resizeFcn(imread(loc));

% Count the number of unique labels
labelCount = numel(unique(faceData.Labels));

% Split the data into training and validation sets
[trainData, valData] = splitEachLabel(faceData, 0.8, 'randomize');

% Load a pre-trained MobileNetV2 network
net = mobilenetv2;

% Create a layer graph from the network
lgraph = layerGraph(net);

% Find the layer to replace
oldLayer = lgraph.Layers(end-2);
newFCLayer = fullyConnectedLayer(labelCount, 'Name', 'new_fc', 'WeightLearnRateFactor', 10, 'BiasLearnRateFactor', 10);

% Replace the fully connected layer with the new one
lgraph = replaceLayer(lgraph, oldLayer.Name, newFCLayer);

% Define the new classification layers
newLayers = [
    softmaxLayer('Name', 'new_softmax')
    classificationLayer('Name', 'new_classoutput')];

% Connect the new layers to the network
lgraph = addLayers(lgraph, newLayers);
lgraph = connectLayers(lgraph, 'new_fc', 'new_softmax');

% Set training options
options = trainingOptions('sgdm', ...
    'MiniBatchSize', 10, ...
    'MaxEpochs', 10, ...
    'InitialLearnRate', 0.001, ...
    'ValidationData', valData, ...
    'ValidationFrequency', 10, ...
    'Verbose', false, ...
    'Plots', 'training-progress');

% Train the network
netTransfer = trainNetwork(trainData, lgraph, options);

% Save the trained network
save('trainedFaceNet.mat', 'netTransfer');