AlexNet网络对已打好标签的四个文件夹中的图片进行分类并对卷积操作提取到的特征可视化具体matlab

代码实现如下：

首先，导入需要用到的工具包和数据集：

% 导入工具包
addpath('./matconvnet-1.0-beta25/matlab');
addpath('./vlfeat-0.9.21/toolbox');
vl_setup;

% 导入数据集
imds = imageDatastore('./data', 'IncludeSubfolders',true, 'LabelSource','foldernames');
[imdsTrain, imdsValidation] = splitEachLabel(imds, 0.7, 'randomized');

然后，定义AlexNet网络结构：

% 定义AlexNet网络结构
net = alexnet;

接着，对训练集数据进行预处理：

% 对训练集数据进行预处理
inputSize = net.Layers(1).InputSize;
augmentedTraining = augmentedImageDatastore(inputSize(1:2), imdsTrain);
augmentedValidation = augmentedImageDatastore(inputSize(1:2), imdsValidation);

然后，对网络进行微调：

% 对网络进行微调
layersTransfer = net.Layers(1:end-3);
numClasses = numel(categories(imdsTrain.Labels));

layers = [
    layersTransfer
    fullyConnectedLayer(numClasses,'WeightLearnRateFactor',20,'BiasLearnRateFactor',20)
    softmaxLayer
    classificationLayer];

options = trainingOptions('sgdm', ...
    'MiniBatchSize',10, ...
    'MaxEpochs',6, ...
    'InitialLearnRate',1e-4, ...
    'Shuffle','every-epoch', ...
    'ValidationData',augmentedValidation, ...
    'ValidationFrequency',3, ...
    'Verbose',false, ...
    'Plots','training-progress');

netTransfer = trainNetwork(augmentedTraining,layers,options);

接着，对测试集数据进行分类：

% 对测试集数据进行分类
imdsTest = imageDatastore('./test', 'IncludeSubfolders',true, 'LabelSource','foldernames');
augmentedTest = augmentedImageDatastore(inputSize(1:2), imdsTest);

predictedLabels = classify(netTransfer,augmentedTest);

最后，对卷积操作提取到的特征进行可视化：

% 对卷积操作提取到的特征进行可视化
I = imread('./test/1/Babyface_0001.jpg');
figure;
imshow(I);

act1 = activations(netTransfer,I,'conv1');
sz = size(act1);
act1 = reshape(act1,[sz(1) sz(2) 1 sz(3)]);
I = imtile(mat2gray(act1),'GridSize',[8 12]);
figure;
imshow(I);

act2 = activations(netTransfer,I,'conv2');
sz = size(act2);
act2 = reshape(act2,[sz(1) sz(2) 1 sz(3)]);
I = imtile(mat2gray(act2),'GridSize',[8 12]);
figure;
imshow(I);

act3 = activations(netTransfer,I,'conv3');
sz = size(act3);
act3 = reshape(act3,[sz(1) sz(2) 1 sz(3)]);
I = imtile(mat2gray(act3),'GridSize',[8 12]);
figure;
imshow(I);

act4 = activations(netTransfer,I,'conv4');
sz = size(act4);
act4 = reshape(act4,[sz(1) sz(2) 1 sz(3)]);
I = imtile(mat2gray(act4),'GridSize',[8 12]);
figure;
imshow(I);

act5 = activations(netTransfer,I,'conv5');
sz = size(act5);
act5 = reshape(act5,[sz(1) sz(2) 1 sz(3)]);
I = imtile(mat2gray(act5),'GridSize',[8 12]);
figure;
imshow(I);

以上代码实现了AlexNet网络对已打好标签的四个文件夹中的图片进行分类，并对卷积操作提取到的特征进行可视化。