Error using activations functions in matlab

Hi,

I trying to replace the pre-trained CNN fully connected layer with mine 'new_fc'.

imds = imageDatastore('MerchData', 'IncludeSubfolders',true, 'LabelSource','foldernames'); 
[imdsTrain,imdsValidation] = splitEachLabel(imds,0.7);

testnet = resnet18
inputSize = testnet.Layers(1).InputSize

augimdsTrain = augmentedImageDatastore(inputSize(1:2),imdsTrain)
augimdsValidation = augmentedImageDatastore(inputSize(1:2),imdsValidation)



if isa(learnableLayer,'nnet.cnn.layer.FullyConnectedLayer')
    newLearnableLayer = fullyConnectedLayer(7, ...
        'Name','new_fc', ...
        'WeightLearnRateFactor',10, ...
        'BiasLearnRateFactor',10);
    
elseif isa(learnableLayer,'nnet.cnn.layer.Convolution2DLayer')
    newLearnableLayer = convolution2dLayer(1,7, ...
        'Name','new_conv', ...
        'WeightLearnRateFactor',10, ...
        'BiasLearnRateFactor',10);
end

lgraph = replaceLayer(lgraph,learnableLayer.Name,newLearnableLayer)

newClassLayer = classificationLayer('Name','new_classoutput')
lgraph = replaceLayer(lgraph,classLayer.Name,newClassLayer)

layer = 'new_fc';
activations(lgraph,augimdsTrain,layer,'OutputAs','rows')

The error message that i got is as below.

"check for incorrect argument data type or missing argument in call to function 'activations'.

Hope someone could help on this. Thank you in advance.

Answer 1

[yanqi liu]

yes，sir，if modify net structure，may be need retrain it，then use activations to get feature

so，may be use

featuresTrain = activations(testnet,augimdsTrain,'fc1000','OutputAs','rows')

[Teo]

Hi @yanqi liu,

Thanks for the reply. May i comfirm with you again that i need to retrain it first using "trainNetwork" function then only use the following code?

featuresTrain = activations(lgraph,augimdsTrain,layer,'OutputAs','rows')

or i can directly use the followig code without retrain it as you already provide the solution as followed?

featuresTrain = activations(lgraph,augimdsTrain,'new_fc','OutputAs','rows')

Hope you are able to help on this?

Thank you

[Teo]

Hi @yanqi liu,

i managed to solve the problem. Thank you for your help.