How can I test the network I've trained?

Hello everybody, I have trained a CNN model which is Resnet50 and now I want to test it with different dataset and get accuracy. Anyone has any idea how can I do that? I have created a imageDataStore called imdsTest which has 5 classes and thousands of images in the classes.

Answer 1

[Christopher McCausland]

Hi Ferhat,

If you have trained the model it will save to your workspace as a fancy variable 'model'. You can use this to then test with images it hasn't seen before.

% Lets generate a network -- Yours may look slightly diffrent, thats okay

miniBatchSize  = 128;
validationFrequency = floor(numel(YTrain)/miniBatchSize);
options = trainingOptions('sgdm', ...
    'MiniBatchSize',miniBatchSize, ...
    'MaxEpochs',30, ...
    'InitialLearnRate',1e-3, ...
    'LearnRateSchedule','piecewise', ...
    'LearnRateDropFactor',0.1, ...
    'LearnRateDropPeriod',20, ...
    'Shuffle','every-epoch', ...
    'ValidationData',{XValidation,YValidation}, ...
    'ValidationFrequency',validationFrequency, ...
    'Plots','training-progress', ...
    'Verbose',false);

% Lets train it;

net = trainNetwork(XTrain,YTrain,layers,options);

% Finally lets test our network;

YPredicted = predict(net,imdsTest);

Here is a further example (that I have also borrowed from as I don't have a network to hand).

Kind regards,

Christopher