MATLAB UNET代码

以下是一个简单的 MATLAB UNET 代码示例，用于对图像进行二进制分类。

首先，加载图像并将其拆分为训练和测试数据集：

% Load images
imds = imageDatastore('path/to/images');

% Split images into training and testing sets
[trainImgs, testImgs] = splitEachLabel(imds, 0.7);

然后，定义 UNET 架构并训练模型：

% Define UNET architecture
layers = [
    imageInputLayer([32 32 1])
    convolution2dLayer(3, 16, 'Padding', 'same')
    batchNormalizationLayer
    reluLayer
    
    maxPooling2dLayer(2, 'Stride', 2)
    
    convolution2dLayer(3, 32, 'Padding', 'same')
    batchNormalizationLayer
    reluLayer
    
    maxPooling2dLayer(2, 'Stride', 2)
    
    convolution2dLayer(3, 64, 'Padding', 'same')
    batchNormalizationLayer
    reluLayer
    
    transposedConv2dLayer(2, 64, 'Stride', 2)
    
    concatenationLayer(3)
    
    convolution2dLayer(3, 32, 'Padding', 'same')
    batchNormalizationLayer
    reluLayer
    
    transposedConv2dLayer(2, 32, 'Stride', 2)
    
    concatenationLayer(2)
    
    convolution2dLayer(3, 16, 'Padding', 'same')
    batchNormalizationLayer
    reluLayer
    
    convolution2dLayer(1, 1)
    regressionLayer
];

% Train model
opts = trainingOptions('adam', ...
    'MaxEpochs', 30, ...
    'MiniBatchSize', 32, ...
    'ValidationData', testImgs, ...
    'Plots', 'training-progress');

model = trainNetwork(trainImgs, layers, opts);

最后，使用测试数据集评估模型：

% Evaluate model on test data
predImgs = predict(model, testImgs);
accuracy = sum(predImgs == testImgs.Labels)/numel(testImgs.Labels);