%%?準備工作空間
clc?
clear?all
close?all
%%?導入要訓練的數(shù)據
digitDatasetPath?=?fullfile（‘./‘?‘/HandWrittenDataset/‘）;
imds?=?imageDatastore（digitDatasetPath‘IncludeSubfolders‘true‘LabelSource‘‘foldernames‘）;
%?采用文件夾名稱作為數(shù)據標記，imageDatastore是一種數(shù)據的讀取方式，IncludeSubfolders：是否繼續(xù)讀取子文件夾下的圖像，‘LabelSource‘‘foldernames‘采用子文件的名字對數(shù)據經行標記
%‘ReadFcn‘@mineRF

%?查看imds數(shù)據集圖片個數(shù)
countEachLabel（imds）
numTrainFiles?=?17;%?每一個數(shù)字有22個樣本，取17個樣本作為訓練數(shù)據
[imdsTrainimdsValidation]?=?splitEachLabel（imdsnumTrainFiles‘randomize‘）;%可選參數(shù)指定為‘隨機化’
%?查看圖片的大小，因為涉及到神經網絡的輸入程大小
img=readimage（imds1）;
size（img）

%%?定義卷積神經網絡的結構Lenet5
layers?=?[
%?輸入層
imageInputlayer（[200?200?3]）%200指的是需要訓練的圖片寬高，RGB圖片，通道只有3
%?卷積層：提取底層特征
convolution2dlayer（56‘Padding‘2）
%卷積核大小5*5，提取6種特征映射，Padding在圖片邊緣進行補零操作以保證在這次卷積之后特征映射大小仍是28*28，計算方法為：28-5+4+1
batchNormalizationlayer
relulayer
%下采樣層，核大小為：2*2?平移不變性、防止過度擬合、減小數(shù)據維度
maxPooling2dlayer（2‘stride‘2）

convolution2dlayer（5?16）
batchNormalizationlayer
relulayer

maxPooling2dlayer（2‘stride‘2）

convolution2dlayer（5?120）
batchNormalizationlayer
relulayer
%?最終層
fullyConnectedlayer（9）%通過一個全連接層，將120個特征連接成9個輸出特征
%該層為了分類
softmaxlayer
classificationlayer];

%%?訓練神經網絡
%?設置訓練參數(shù)
options?=?trainingOptions（‘sgdm‘‘maxEpochs‘60?‘ValidationData‘?imdsValidation‘ValidationFrequency‘5‘Verbose‘false‘Plots‘‘training-progress‘）;%?顯示訓練進度。
%sgdm最優(yōu)化的方法，60重復次數(shù)，imdsValidation驗證集，5：驗證的頻率，‘Verbose‘false不顯示中間驗證的結果，training-progress‘顯示訓練進度


%?訓練神經網絡，保存網絡
ZQ=?trainNetwork（imdsTrain?layers?options）;%參數(shù)分別為訓練數(shù)據、神經元設計、訓練參數(shù)
save?‘ZQ.mat‘?ZQ

?屬性????????????大小?????日期????時間???名稱
-----------?---------??----------?-----??----
?????目錄???????????0??2020-04-16?17:43??Design\HandWrittenDataset\
?????目錄???????????0??2020-04-16?17:43??Design\HandWrittenDataset\0\
?????文件????????4364??2020-04-12?07:59??Design\HandWrittenDataset\0\1.jpg
?????文件????????5799??2020-04-12?07:59??Design\HandWrittenDataset\0\10.jpg
?????文件????????6552??2020-04-12?07:59??Design\HandWrittenDataset\0\11.jpg
?????文件????????5723??2020-04-12?07:59??Design\HandWrittenDataset\0\12.jpg
?????文件????????5292??2020-04-12?07:59??Design\HandWrittenDataset\0\13.jpg
?????文件????????5338??2020-04-12?07:59??Design\HandWrittenDataset\0\14.jpg
?????文件????????5156??2020-04-12?07:59??Design\HandWrittenDataset\0\15.jpg
?????文件????????7121??2020-04-12?07:59??Design\HandWrittenDataset\0\16.jpg
?????文件????????6361??2020-04-12?07:59??Design\HandWrittenDataset\0\17.jpg
?????文件????????7426??2020-04-12?07:59??Design\HandWrittenDataset\0\18.jpg
?????文件????????4902??2020-04-12?07:59??Design\HandWrittenDataset\0\19.jpg
?????文件????????5805??2020-04-12?07:59??Design\HandWrittenDataset\0\2.jpg
?????文件????????5990??2020-04-12?07:59??Design\HandWrittenDataset\0\20.jpg
?????文件????????4945??2020-04-12?07:59??Design\HandWrittenDataset\0\21.jpg
?????文件????????5688??2020-04-12?07:59??Design\HandWrittenDataset\0\22.jpg
?????文件????????6052??2020-04-12?07:59??Design\HandWrittenDataset\0\3.jpg
?????文件????????7426??2020-04-12?07:59??Design\HandWrittenDataset\0\4.jpg
?????文件????????4196??2020-04-12?07:59??Design\HandWrittenDataset\0\5.jpg
?????文件????????6161??2020-04-12?07:59??Design\HandWrittenDataset\0\6.jpg
?????文件????????4418??2020-04-12?07:59??Design\HandWrittenDataset\0\7.jpg
?????文件????????6809??2020-04-12?07:59??Design\HandWrittenDataset\0\8.jpg
?????文件????????5286??2020-04-12?07:59??Design\HandWrittenDataset\0\9.jpg
?????目錄???????????0??2020-04-16?17:43??Design\HandWrittenDataset\1\
?????文件????????5255??2020-04-12?07:59??Design\HandWrittenDataset\1\1.jpg
?????文件????????4150??2020-04-12?07:59??Design\HandWrittenDataset\1\10.jpg
?????文件????????6620??2020-04-12?07:59??Design\HandWrittenDataset\1\11.jpg
?????文件????????4569??2020-04-12?07:59??Design\HandWrittenDataset\1\12.jpg
?????文件????????4297??2020-04-12?07:59??Design\HandWrittenDataset\1\13.jpg
?????文件????????4607??2020-04-12?07:59??Design\HandWrittenDataset\1\14.jpg
............此處省略184個文件信息