why do I get the error Undefined function or variable 'net'. Error in testalexnet1 (line 16) trainingFeatures = activation​s(net,trai​ningImages​,layer);

I am trying to create my own image data store and this is my code

trainingFeatures = activations(net,trainingImages,layer);

[Hridya PI]

I just downloded matlab 2017 trial version .and alexnet too. I have some images of indian coastal fishes to identify. So I created ImageDataStore object and started preparing training and testing images. and the error comes here..

trainingFeatures = activations(net,trainingImages,layer);

Undefined function or variable 'net'.

this is the error it shows. What may be the reason for it?

[Walter Roberson]

Please show your code before that. In particular, please show your trainNetwork call.

[Hridya PI]

imds1 = imageDatastore(fullfile(matlabroot,'toolbox','matlab','images','New Folder'),...
'IncludeSubfolders',true,'FileExtensions','.jpg','LabelSource','foldernames')
%data = read(imds)

[Hridya PI]

[trainingImages,testImages] = splitEachLabel(imds1,0.7,'randomized');
numTrainImages = numel(trainingImages.Labels);
idx = randperm(numTrainImages,11);
figure
for i = 1:11
    subplot(4,4,i)
    I = readimage(trainingImages,idx(i));
    imshow(I)
end

layer = 'fc7';
trainingFeatures = activations(net,trainingImages,layer);
testFeatures = activations(net,testImages,layer);
trainingLabels = trainingImages.Labels;
testLabels = testImages.Labels;
classifier = fitcecoc(trainingFeatures,trainingLabels);
predictedLabels = predict(classifier,testFeatures);idx = [1 5 10 15];
figure
for i = 1:numel(idx)
    subplot(2,2,i)
    I = readimage(testImages,idx(i));
    label = predictedLabels(idx(i));
    imshow(I)
    title(char(label))
end

Answer 1

[Walter Roberson]

Considering the name of your file testalexnet1 it appears that you missed

net = alexnet;

https://www.mathworks.com/help/nnet/examples/visualize-activations-of-a-convolutional-neural-network.html

[Hridya PI]

ok..thank you.. let me try

Answer 2

[Hridya PI]

Error using SeriesNetwork/activations (line 794)
'OutputAs' must be 'channels' to use activations on images larger than the network's imageInputLayer.InputSize.

Error in testalexnet1 (line 17)
trainingFeatures = activations(net,trainingImages,layer);

now this is the error .

i changed the line of code as

trainingFeatures = activations(net,trainingImages,layer,'outputAs','channels');

is this right?

[Walter Roberson]

It might be okay; I am not familiar with that function.

[Hridya PI]

trainingFeatures = activations(net,trainingImages,'fc7','OutputAs','channels');
testFeatures = activations(net,testImages,'fc7','OutputAs','channels');

this code corrected the error.

[Hridya PI]

 "Error using classreg.learning.FullClassificationRegressionModel.prepareDataCR (line 192)
X must be a numeric matrix."

What may be the possible reason for this error?

[Walter Roberson]

Possibly you could get that error if you were passing in a 3D array instead of 2D. 3D would occur for a color image.

[Hridya PI]

Thanks a lot... and also some images exceeded the size of input layer. 227×227×3

[Hridya PI]

Now its fixed

[Hridya PI]

Now, I completed the Alexnet Training with my dataset. How I can input an image and make the network predict it? I tried the code below.Butb it takes tha classnames of the pretrained network.Not the classnames that I newly created.

   img= imread('D:\as.jpg');
  img = imresize(I,[227 227]);
  label = classify(net,img)
  figure
  imshow(img)
  title(char(label))

and also the code

net.Layers(end).ClassNames(1:6)

gives the classnames of pretrained network only. What should I do to give an input to the net. Please help