LSTM input size mismatch

Hi,

I am meeting a trouble when I want to build a network with CNN and LSTM. Basically, I learn and follow from this link: https://uk.mathworks.com/help/deeplearning/ug/classify-videos-using-deep-learning.html. my dataset has 2000 samples (each sample contains 40 points), when I test my custom network, there exists an error on the LSTM layer:

Layer 'lstm': Input size mismatch. Size of input to this layer is different from the expected input size.

Inputs to this layer:

from layer 'flatten' (output size 100)

Attached please find the analysis of my network architecture

Answer 1

[Srivardhan Gadila]

With the following code I was able to construct layerGraph with no errors. Make sure you define the lstmLayer with right input arguments.

inputSize = [1 40 1];
filterSize = [1 3];
numFilters = 5;
numHiddenUnits = 25;
numClasses = 2;

layers = [ ...
    sequenceInputLayer(inputSize,'Name','input')
    
    sequenceFoldingLayer('Name','fold')
    
    convolution2dLayer(filterSize,numFilters,'Name','conv','stride',[1 1],'padding','same')

    maxPooling2dLayer([1 2],'padding',[0 0 0 0],'Name','maxpool','Stride',[1 2])
    
    sequenceUnfoldingLayer('Name','unfold')
    flattenLayer('Name','flatten')
    
    lstmLayer(numHiddenUnits,'OutputMode','last','Name','lstm')
    
    fullyConnectedLayer(numClasses, 'Name','fc')
    softmaxLayer('Name','softmax')
    classificationLayer('Name','classification')];
lgraph = layerGraph(layers);
lgraph = connectLayers(lgraph,'fold/miniBatchSize','unfold/miniBatchSize');
analyzeNetwork(lgraph)