複数の入力層を持つC​NNにおいて加算層を​持つ事前学習済みネッ​トワークを使用する際​に発生するエラーにつ​いて

https://jp.mathworks.com/matlabcentral/fileexchange/74760-image-classification-using-cnn-with-multi-input-cnn

上記の手法を参考に2種類の画像データセットを用いたCNNの作成を試みているのですが，ネットワークのアーキテクチャを

事前学習済みネットワークに置き換え学習したところResNet-18やResNet-50、GoogleNetなど加算層を持つネットワークでエラーが発生しました．

エラー文から未接続の入力があると分かりましたが改善方法が分かりませんでした。

以下にプログラムの一部とエラー文を示します。このエラーの改善方法を教えていただきたいです。

よろしくお願いいたします。

%データセット
auimdsTrain=augmentedImageDatastore([224,224],imds,'OutputSizeMode','centercrop');
auimdsValidation=augmentedImageDatastore([224 224],imds3,'OutputSizeMode','centercrop');
auimdsTrain2=augmentedImageDatastore([224,224],imds2,'OutputSizeMode','centercrop');
auimdsValidation2=augmentedImageDatastore([224,224],imds4,'OutputSizeMode','centercrop');
auimdsTest=augmentedImageDatastore([224,224],imds5,'OutputSizeMode','centercrop');
auimdsTest2=augmentedImageDatastore([224,224],imds6,'OutputSizeMode','centercrop');

label1=imds.Labels;
label2=imds3.Labels;
label3=imds5.Labels;
[XTrainUpper,XTrainBottom,XValidUpper,XValidBottom,XTestUpper,XTestBottom,YTrain,YValid,YTest]=partitionData(auimdsTrain,auimdsTrain2,auimdsValidation,auimdsValidation2,auimdsTest,auimdsTest2,label1,label2,label3);
classes=categories(YTrain);
numClasses=numel(classes);

%ネットワーク
net=resnet18;
net=layerGraph(net);
layersTransfer=net.Layers(2:end-3);
layersTransfer2=net.Layers(3:end-3);
numHiddenDimension=142; % speficy the dimension of the hidden layer
layers = createSimpleLayer(XTrainUpper,numHiddenDimension,layersTransfer);
layers2 = createSimpleLayer2(XTrainBottom,numHiddenDimension,layersTransfer2);

layers2=renameLayer(layers2,'_2');
layersAdd=[fullyConnectedLayer(numHiddenDimension,'Name','fcAdd1')
    fullyConnectedLayer(numClasses,'Name','fcAdd2')];
layersRemoved=[layers(1:end);concatenationLayer(1,2,'Name','cat');layersAdd];
lgraphAggregated = addLayers(layerGraph(layersRemoved),layers2(1:end));
lgraphAggregated = connectLayers(lgraphAggregated,'fc_2','cat/in2');

dlnet = dlnetwork(lgraphAggregated); % A dlnetwork object enables support for custom training loops using automatic differentiation

%アーキテクチャ定義
function layers=createSimpleLayer(XTrainData_4D,numHiddenDimension,layersTransfer)
    layers = [
        imageInputLayer([224 224 3],"Name","imageinput","Mean",mean(XTrainData_4D,'all'))
        layersTransfer
        fullyConnectedLayer(numHiddenDimension,"Name","fc")];
end
function layers=createSimpleLayer2(XTrainData_4D,numHiddenDimension,layersTransfer2)
    layers = [
        imageInputLayer([224 224 1],"Name","imageinput","Mean",mean(XTrainData_4D,'all'))
        convolution2dLayer([7 7],64,"Stride",[2 2],"DilationFactor",[1 1],"Padding",[3,3,3,3],"PaddingValue",0,"BiasLearnRateFactor",0,"WeightL2Factor",1,"Name","conv_1","Padding","same")
        layersTransfer2
        fullyConnectedLayer(numHiddenDimension,"Name","fc")];
end
%パーティションデータ
function [XTrainUpper,XTrainBottom,XValidUpper,XValidBottom,XTestUpper,XTestBottom,YTrain,YValid,YTest]=partitionData(auimdsTrain,auimdsTrain2,auimdsValidation,auimdsValidation2,auimdsTest,auimdsTest2,label1,label2,label3)

XTrainUpper=readall(auimdsTrain);
XTrainUpper=cat(4,XTrainUpper.input{:});
XTrainBottom=readall(auimdsTrain2);
XTrainBottom=cat(4,XTrainBottom.input{:});
XValidUpper=readall(auimdsValidation);
XValidUpper=cat(4,XValidUpper.input{:});
XValidBottom=readall(auimdsValidation2);
XValidBottom=cat(4,XValidBottom.input{:});
XTestUpper=readall(auimdsTest);
XTestUpper=cat(4,XTestUpper.input{:});
XTestBottom=readall(auimdsTest2);
XTestBottom=cat(4,XTestBottom.input{:});

YTrain=label1;
YValid=label2;
YTest=label3;
end
%レイヤー名変更
function layers=renameLayer(layers,char)
   for i=1:numel(layers)
       layers(i).Name=[layers(i).Name,char];
   end
end

使い方によるエラー dlnetwork/initialize

無効なネットワーク。

エラー: dlnetwork (行 218)

net = initialize(net, dlX{:});

エラー: test1 (行 32)

dlnet = dlnetwork(lgraphAggregated); % A dlnetwork object enables support for custom training loops using automatic differentiation

原因:

Example inputs: Incorrect number of example network inputs. 0 example network inputs provided but network has 18 inputs including 16 unconnected layer

inputs.

層 'res2a': Unconnected input. Each input must be connected to input data or to the output of another layer.

未接続の入力を検出しました:

入力 'in2'

層 'res2a_2': Unconnected input. Each input must be connected to input data or to the output of another layer.

未接続の入力を検出しました:

入力 'in2'

層 'res2b': Unconnected input. Each input must be connected to input data or to the output of another layer.

未接続の入力を検出しました:

入力 'in2'

層 'res2b_2': Unconnected input. Each input must be connected to input data or to the output of another layer.

未接続の入力を検出しました:

入力 'in2'

層 'res3a': Unconnected input. Each input must be connected to input data or to the output of another layer.

未接続の入力を検出しました:

入力 'in2'

層 'res3a_2': Unconnected input. Each input must be connected to input data or to the output of another layer.

未接続の入力を検出しました:

入力 'in2'

層 'res3b': Unconnected input. Each input must be connected to input data or to the output of another layer.

未接続の入力を検出しました:

入力 'in2'

層 'res3b_2': Unconnected input. Each input must be connected to input data or to the output of another layer.

未接続の入力を検出しました:

入力 'in2'

層 'res4a': Unconnected input. Each input must be connected to input data or to the output of another layer.

未接続の入力を検出しました:

入力 'in2'

層 'res4a_2': Unconnected input. Each input must be connected to input data or to the output of another layer.

未接続の入力を検出しました:

入力 'in2'

層 'res4b': Unconnected input. Each input must be connected to input data or to the output of another layer.

未接続の入力を検出しました:

入力 'in2'

層 'res4b_2': Unconnected input. Each input must be connected to input data or to the output of another layer.

未接続の入力を検出しました:

入力 'in2'

層 'res5a': Unconnected input. Each input must be connected to input data or to the output of another layer.

未接続の入力を検出しました:

入力 'in2'

層 'res5a_2': Unconnected input. Each input must be connected to input data or to the output of another layer.

未接続の入力を検出しました:

入力 'in2'

層 'res5b': Unconnected input. Each input must be connected to input data or to the output of another layer.

未接続の入力を検出しました:

入力 'in2'

層 'res5b_2': Unconnected input. Each input must be connected to input data or to the output of another layer.

未接続の入力を検出しました:

入力 'in2'