%% train_1transfer googlenet SqueezeNet aynı parametreler save
%%
  imdsTrain = imageDatastore('800', 'IncludeSubfolders',true, 'LabelSource','foldernames');
 % [imdsTrain,imdsValidation] = splitEachLabel(imds,0.9,'randomized');
% save data.mat imds imdsTrain imdsValidation
 % load data.mat;

 %test
 net=netTransfer;
   timds = imageDatastore('test800', 'IncludeSubfolders',true, 'LabelSource','foldernames');
  inputSize = net.Layers(1).InputSize;
ttimds = augmentedImageDatastore(inputSize(1:2),timds);
   [YPred,scores] = classify(netTransfer,ttimds);imdsTrain = imageDatastore('800', 'IncludeSubfolders',true, 'LabelSource','foldernames');

   confMat = confusionmat(timds.Labels, YPred)
   
cm = confusionchart(confMat);
fig = figure;
cm = confusionchart(confMat,'ColumnSummary','column-normalized');

writematrix(info.TrainingAccuracy,'info.xls','Sheet','shufflenet','Range','B6');% excel 
writematrix(info.TrainingLoss,'info.xls','Sheet','shufflenet','Range','B7');% excel 


%% NET
%%  imdsTrain=imds;
% net = inceptionv3;
% net=resnet18;
% net=vgg19;
% net=inceptionv3;
% net=mobilenetv2;
% net=densenet201;
% net=resnet101;
% net=inceptionresnetv2;
% net=efficientnetb0;
% net=shufflenet;

lgraph = layerGraph(net);
[learnableLayer,classLayer] = findLayersToReplace(lgraph);
numClasses = numel(categories(imdsTrain.Labels));
if isa(learnableLayer,'nnet.cnn.layer.FullyConnectedLayer')
    newLearnableLayer = fullyConnectedLayer(numClasses, ...
        'Name','new_fc', ...
        'WeightLearnRateFactor',10, ...
        'BiasLearnRateFactor',10);
elseif isa(learnableLayer,'nnet.cnn.layer.Convolution2DLayer')
    newLearnableLayer = convolution2dLayer(1,numClasses, ...
        'Name','new_conv', ...
        'WeightLearnRateFactor',10, ...
        'BiasLearnRateFactor',10);
end
lgraph = replaceLayer(lgraph,learnableLayer.Name,newLearnableLayer);
newClassLayer = classificationLayer('Name','new_classoutput');
lgraph = replaceLayer(lgraph,classLayer.Name,newClassLayer);


%% train
%% no auq
options = trainingOptions('adam', ...
    'MiniBatchSize',64, ... % 256  2nin katı 4 veya 8 16 2, 4, 8, 16, 32, … 512 vb.
    'MaxEpochs',2, ...
    'InitialLearnRate',1e-4, ...
    'Shuffle','every-epoch', ...
    'Verbose',false, ...
    'Plots','training-progress',...
    'LearnRateSchedule', "piecewise");
% 'LearnRateDropPeriod', 6

inputSize = net.Layers(1).InputSize;
augimdsTrain = augmentedImageDatastore(inputSize(1:2),imdsTrain);
tic;
[netTransfer,info]=trainNetwork(augimdsTrain,lgraph,options);
toc
% input image 224 224 ayarlamak için augmentedImageDatastore
% imdsValidation = augmentedImageDatastore(inputSize(1:2),imdsValidation);

% save resnet101epoch2size64.mat netTransfer info
% save shufflenetepoch100size64.mat netTransfer info
 epoch2 32minesize 1 saat
%%
%%
%%
%% imageAugmenter
%%
inputSize = net.Layers(1).InputSize;
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);




%% first googlenet fullyConnectedLayer *******************************************
%%
net = googlenet; 
%deepNetworkDesigner(net)
inputSize = net.Layers(1).InputSize;
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);

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);

options = trainingOptions('adam', ...
    'MiniBatchSize',128, ... % 256  2nin katı 4 veya 8 16 2, 4, 8, 16, 32, … 512 vb.
    'MaxEpochs',8, ...
    'InitialLearnRate',1e-4, ...
    'Shuffle','every-epoch', ...
    'ValidationData',augimdsValidation, ...
    'ValidationFrequency',64, ... % 30 yap
    'Verbose',false, ...
    'Plots','training-progress');

%netTransfer = trainNetwork(augimdsTrain,lgraph,options);
[netTransfer,info]=trainNetwork(augimdsTrain,lgraph,options);

% save netgooglepoch8size128.mat netTransfer info

%% ************************************************
%%
%% To retrain SqueezeNet conv10' and 'ClassificationLayer_predictions'
%%
load data.mat;

net = squeezenet;
inputSize = net.Layers(1).InputSize;
lgraph = layerGraph(net); 
numClasses = numel(categories(imdsTrain.Labels));
newConvLayer =  convolution2dLayer([1, 1],numClasses,'WeightLearnRateFactor',10,...
    'BiasLearnRateFactor',10,...
    "Name",'new_conv');
lgraph = replaceLayer(lgraph,'conv10',newConvLayer);
newClassificatonLayer = classificationLayer('Name','new_classoutput');
lgraph = replaceLayer(lgraph,'ClassificationLayer_predictions',newClassificatonLayer);

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);

options = trainingOptions('adam', ...
    'MiniBatchSize',128, ... % 256  2nin katı 4 veya 8 16 2, 4, 8, 16, 32, … 512 vb.
    'MaxEpochs',10, ...
    'InitialLearnRate',1e-4, ...
    'Shuffle','every-epoch', ...
    'ValidationData',augimdsValidation, ...
    'ValidationFrequency',30, ... % 30 yap
    'Verbose',false, ...
    'Plots','training-progress');

netTransfer = trainNetwork(augimdsTrain,lgraph,options);
 [netTransfer,info]=trainNetwork(augimdsTrain,lgraph,options);

% save netsqueezenetepoch8size128.mat netTransfer info

%%
% miniBatchSize = 128;
% learnRate = 0.1*miniBatchSize/128;
options = trainingOptions("sgdm", ...
    'InitialLearnRate', 1e-4,...
    'MaxEpochs', 60, ...% 30
    'MiniBatchSize', 128, ... % 256  2nin katı 4 veya 8 16 2, 4, 8, 16, 32, … 512 vb.
    'ValidationFrequency',30, ... % 30 yap
    'Shuffle','every-epoch', ...
    'Plots','training-progress',...
    'Verbose',false, ...
    'ValidationData',augimdsValidation, ...
    'LearnRateSchedule', "piecewise");

[netTransfer,info]=trainNetwork(augimdsTrain,lgraph,options);

% save netsqueezenet.mat netTransfer info