how does it possible in convnet, High accuracy in both of validation and test the same has 97% with error loss function 0.01

2 vues (au cours des 30 derniers jours)
i really need an experience help . i feel misunderstand about convolution neural network i have a medical image data set and i want to use a pretrained alexnet so i wrote this code in matlab ( very simple code) myimages='D:\ ';
allimages=imageDatastore(myimages,'IncludeSubfolders',true,'FileExtensions','.png','LabelSource','foldernames');
allimages.ReadFcn= @(filename)readAndPreprocessImage(filename);
[trainingimages,valDigitData,testimage]=splitEachLabel(allimages,0.4,0.2,'randomize');
net= alexnet; layersTransfer = net.Layers(2:end-3);
here i just used layers from 2
%% my layers Layers =[... imageInputLayer([227 227 3],'Name','input') layersTransfer
fullyConnectedLayer(2,'Name','FC_3','WeightLearnRateFactor',20,'BiasLearnRateFactor',20) softmaxLayer('Name','prob') classificationLayer('Name','coutput')];
Layers(23).Weights = randn([2 9216]) * 0.0001;
Layers(23).Bias = randn([2 1])*0.0001 + 1;
opts=trainingOptions('sgdm','Initiallearnrate',0.0001,'maxEpoch',5, 'Minibatchsize',256,'L2Regularization',0.0001, 'Plots','training-progress','ValidationData',valDigitData,'ValidationFrequency',50,'LearnRateSchedule','piecewise','LearnRateDropFactor',0.1,'LearnRateDropPeriod', 1);
%% Re_train the network [trainedNet,traininfo] = trainNetwork(trainingimages,Layers,opts);
%% Classify test Images [predictedlabels,error_test]=classify(trainedNet,testimage);
accuracy= mean(predictedlabels== testimage.Labels);
my concern about i got the same accuracy 97% in valid and test how it can be and error is 0.01 . is this code correct (good work)does the code indeed classify my medical images or ( the alexnet images ) or what ? did i misunderstand something?

Réponses (0)

Catégories

En savoir plus sur Image Data Workflows dans Help Center et File Exchange

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by