Error using nnet.cnn.L​ayerGraph>​iValidateL​ayerName (line 654) Layer 'Classific​ationLayer​_predictio​ns' does not exist. Error in nnet.cnn.L​ayerGraph/​replaceLay​er (line 397)

clc

clear all

outputFolder=fullfile('recycle101');

rootFolder=fullfile(outputFolder,'recycle');

categories={'Aluminium Can','PET Bottles','Drink Carton Box'};

imds=imageDatastore(fullfile(rootFolder,categories),'LabelSource','foldernames');

tbl=countEachLabel(imds)

minSetCount=min(tbl{:,2});

[imdsTrain,imdsValidation] = splitEachLabel(imds,0.7,'randomized');

countEachLabel(imds);

%randomly choose file for aluminium can, PET bottles, and drink carton box

AluminiumCan=find(imds.Labels=='Aluminium Can',1);

PETBottles=find(imds.Labels=='PET Bottles',1);

DrinkCartonBox=find(imds.Labels=='Drink Carton Box',1);

%plot image that was pick randomly

figure

subplot(2,2,1);

imshow(readimage(imds,AluminiumCan));

subplot(2,2,2);

imshow(readimage(imds,PETBottles));

subplot(2,2,3);

imshow(readimage(imds,DrinkCartonBox));

%Load pre-trained network

net = resnet50;

analyzeNetwork(net)

numClasses = numel(categories(imdsTrain.Labels));

lgraph = layerGraph(net);

%Replace the classification layers for new task

newFCLayer = fullyConnectedLayer(3,'Name','new_fc','WeightLearnRateFactor',10,'BiasLearnRateFactor',10);

lgraph = replaceLayer(lgraph,'fc1000',newFCLayer);

newClassLayer = classificationLayer('Name','new_classoutput');

lgraph = replaceLayer(lgraph,'ClassificationLayer_predictions',newClassLayer);

Answer 1

[yanqi liu]

may be use

lgraph = replaceLayer(lgraph,'ClassificationLayer_fc1000',newClassLayer);

[TAN HOR YAN]

Owhh is works!!

Thank you so much!

May I know what is the different between ClassificationLayer_fc1000 and ClassificationLayer_predictions?

[yanqi liu]

yes，sir，when we see the layers，we can find the Classification Output name，such as

net = resnet50('Weights','none');
lys = net.Layers;
lys(end-3:end)

ans = 
  4×1 Layer array with layers:

     1   'avg_pool'                     2-D Global Average Pooling   2-D global average pooling
     2   'fc1000'                       Fully Connected              1000 fully connected layer
     3   'fc1000_softmax'               Softmax                      softmax
     4   'ClassificationLayer_fc1000'   Classification Output        crossentropyex

[TAN HOR YAN]

I see. Thank you so much!

[TAN HOR YAN]

Can I email my coding for you to check and give me some advice?

[yanqi liu]

yes，sir，may be send me the file and some information to do analysis，the email address is

[e-mail]

[TAN HOR YAN]

done email. Please check mailbox.

Thank you so much!