An error using trainNetwork. Invalid training data. The output size (1311) of the last layer does not match the number of classes of the responses (20).

Question

Afiq Anang 2023-11-14

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2046947-an-error-using-trainnetwork-invalid-training-data-the-output-size-1311-of-the-last-layer-does-no

回答： Gagan Agarwal 2023-11-20

rootFolder = fullfile('');
categories = {'A', 'B', 'C', 'D', 'E',...
    'F', 'G', 'H', 'I', 'J',...
    'K', 'L', 'M', 'N', 'O',...
    'P', 'Q', 'R', 'S', 'T'};
handles.categories = categories;
%% Store categories into imds
imds = imageDatastore(fullfile(rootFolder, categories), 'LabelSource', 'foldernames');
handles.imds = imds;
[imdsTrain,imdsValidation] = splitEachLabel(imds,0.7,'randomized');
%% Load net
net = googlenet;
deepNetworkDesigner(net)
inputSize = net.Layers(1).InputSize;
%% Replace final layer
lgraph = layerGraph(net); 
numClasses = numel(categories(imdsTrain.Labels));
newLearnableLayer = fullyConnectedLayer(numClasses, ...
    'Name','new_fc', ...
    'WeightLearnRateFactor',10, ...
    'BiasLearnRateFactor',10);
lgraph = replaceLayer(lgraph,'loss3-classifier',newLearnableLayer);
newClassLayer = classificationLayer('Name','new_classoutput');
lgraph = replaceLayer(lgraph,'output',newClassLayer);
%% Train network
pixelRange = [-30 30];
imageAugmenter = imageDataAugmenter( ...
    'RandXReflection',true, ...
    'RandXTranslation',pixelRange, ...
    'RandYTranslation',pixelRange);
augimdsTrain = augmentedImageDatastore(inputSize(1:2),imdsTrain, ...
    'DataAugmentation',imageAugmenter);
augimdsValidation = augmentedImageDatastore(inputSize(1:2),imdsValidation);
%% Training Option
options = trainingOptions('sgdm', ...
    'MiniBatchSize',10, ...
    'MaxEpochs',6, ...
    'InitialLearnRate',1e-4, ...
    'Shuffle','every-epoch', ...
    'ValidationData',augimdsValidation, ...
    'ValidationFrequency',3, ...
    'Verbose',false, ...
    'Plots','training-progress');
%% Training Progress Plot
netTransfer = trainNetwork(augimdsTrain,lgraph,options);

Error using trainNetwork

Invalid training data. The output size (1311) of the last layer does not match the number of classes of the responses (20).

Error in test (line 55)

netTransfer = trainNetwork(augimdsTrain,lgraph,options);

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Matt J 2023-11-14

You need to explain why you think the error message is invalid. We can't see why.

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Gagan Agarwal 2023-11-20

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2046947-an-error-using-trainnetwork-invalid-training-data-the-output-size-1311-of-the-last-layer-does-no#answer_1356307

Hi Afiq,

I understand that you are encountering an error due to a mismatch between the output layer and the number of classes in the response. This issue occurs when the number of output nodes in the neural network exceeds the number of output classes. To resolve this error, you can adjust the number of output nodes in the network to match the number of output classes. You can achieve this by utilizing the 'outputSize' attribute of the 'fullyConnectedLayer' function.

For additional information please refer to the following documentation link:

https://www.mathworks.com/help/deeplearning/ref/nnet.cnn.layer.fullyconnectedlayer.html

I hope it helps!