CNN-LSTM in MATLAB

I'm trying to implement a CNN + LSTM, but I have an error:

Invalid training data. If the network outputs sequences, then regression responses must be a cell array

of numeric sequences, or a single numeric sequence.

or

Invalid training data. Responses must be a matrix of numeric responses, or a N-by-1 cell array of

sequences, where N is the number of sequences. The feature dimension of all sequences must be the same.

or

Invalid training data. X and Y must have the same number of observations.

what is my mistake?

my code is:

inputSize = [20 25 3];
filterSize = 5;
numFilters = 20;
numHiddenUnits = 5;
numResponses = 1;
layers = [ ...
    sequenceInputLayer(inputSize,'Name','input')
    
    sequenceFoldingLayer('Name','fold')
    
    convolution2dLayer(filterSize,numFilters,'Padding','same','Name','conv')
    batchNormalizationLayer('Name','bn')
    reluLayer('Name','relu')
    
    sequenceUnfoldingLayer('Name','unfold')
    flattenLayer('Name','flatten')
    
    lstmLayer(numHiddenUnits,'OutputMode','sequence','Name','lstm')
    
    fullyConnectedLayer(1, 'Name','fc')
    regressionLayer('Name','regression')];
lgraph = layerGraph(layers);
lgraph = connectLayers(lgraph,'fold/miniBatchSize','unfold/miniBatchSize');


options = trainingOptions('adam', ...
    'MaxEpochs',100, ...
    'GradientThreshold',1, ...
    'InitialLearnRate',0.005, ...
    'LearnRateSchedule','piecewise', ...
    'LearnRateDropPeriod',125, ...
    'LearnRateDropFactor',0.2, ...
    'Verbose',false, ...
    'Plots','training-progress');


net = trainNetwork(input_train,target_train,lgraph,options);

My input_train is a 4-D double that its size is (20,25,3,58) in other words I have 3 sets of images that each set containes 58 set of images with size of 20x25

and when I set target_train a matrix of (58,500) or a 4-D double that its size is (1,500,1,58) or a 4-D double that its size is (1,20,25,58) and

fully connected layer to fullyConnectedLayer(500, 'Name','fc') or fullyConnectedLayer(1, 'Name','fc') or fullyConnectedLayer([20 25], 'Name','fc') the error is:

in all cases the error is:

"Invalid training data. Sequence responses must have the same sequence length as the corresponding

predictors."

[HONGYI LI]

I have the same issue, do you have solved? thanks

[nazanin behfar]

Unfortunately no.

Answer 1

[H W]

添加层分支

将网络分支添加到层次图中。每个分支均为一个线性层组。

tempLayers = [

sequenceInputLayer([20 25 3],"Name","sequence")

sequenceFoldingLayer("Name","seqfold")];

lgraph = addLayers(lgraph,tempLayers);

tempLayers = [

convolution2dLayer([5 5],32,"Name","conv","Padding","same")

batchNormalizationLayer("Name","batchnorm")

reluLayer("Name","relu")];

lgraph = addLayers(lgraph,tempLayers);

tempLayers = [

sequenceUnfoldingLayer("Name","sequnfold")

flattenLayer("Name","flatten")

lstmLayer(5,"Name","lstm")

fullyConnectedLayer(1,"Name","fc")

regressionLayer("Name","regressionoutput")];

lgraph = addLayers(lgraph,tempLayers);

% 清理辅助变量

clear tempLayers;

连接层分支

连接网络的所有分支以创建网络图。

lgraph = connectLayers(lgraph,"seqfold/out","conv");

lgraph = connectLayers(lgraph,"seqfold/miniBatchSize","sequnfold/miniBatchSize");

lgraph = connectLayers(lgraph,"relu","sequnfold/in");