public void CreateAndCompileModel3()
{
Console.WriteLine("Creating model");
GlobalRandom.InitializeRandom();
int imgSize = 75;
ReluActivation reluActivation = new ReluActivation();
SoftmaxActivation softmaxActivation = new SoftmaxActivation();
model = new ConvolutionalNeuralNetwork(imgSize, "rgb");
model.Add(new ConvolutionalLayer(5, 5, reluActivation, "valid"));
model.Add(new MaxPoolingLayer());
model.Add(new ConvolutionalLayer(5, 3, reluActivation, "valid"));
model.Add(new MaxPoolingLayer());
model.Add(new DropoutLayer(0.2));
model.Add(new FlattenLayer());
model.Add(new DropoutLayer(0.5));
model.Add(new DenseLayer(26, softmaxActivation));
Console.WriteLine("Model created");
model.Compile();
Console.WriteLine("Model compiled");
}
public MnistCNNGUI()
{
InitializeComponent();
string trainImgPath = "res/train-images.idx3-ubyte"
, trainLblPath = "res/train-labels.idx1-ubyte"
, testImgPath = "res/t10k-images.idx3-ubyte"
, testLblPath = "res/t10k-labels.idx1-ubyte";
trainingImages = MnistImage.ProcessMNISTFile(trainImgPath, trainLblPath);
testImages = MnistImage.ProcessMNISTFile(testImgPath, testLblPath);
mnistNetwork = new ConvolutionalNeuralNetwork(new ConvolutionalNeuralNetworkProps(28, 28, 90, 10, 6, 5, 5, 2, 2));
mnistNetwork.UploadTrainingSet(trainingImages);
mnistNetwork.UploadTestSet(testImages);
mnistNetwork.NumEpochs = 1;
//mnistNetwork.TestTrainingSet();
mnistNetwork.Test(true);
UpdateSensitivity();
selectedImageSet = testImages;
LoadImage(0);
UseTrainingSetCheckBox_CheckedChanged(null, null);
NetworkWorker = new BackgroundWorker();
NetworkWorker.WorkerReportsProgress = true;
NetworkWorker.DoWork += TrainNetwork;
NetworkWorker.ProgressChanged += (object sender, ProgressChangedEventArgs e) =>
{
int percent = e.ProgressPercentage;
TrainingProgressLabel.Text = $"Network Progress: {percent}%";
TrainingProgressBar.Value = percent;
};
NetworkWorker.RunWorkerCompleted += (object sender, RunWorkerCompletedEventArgs e) =>
{
UpdateSensitivity();
SetNetworkActive(true);
TrainingProgressLabel.Text = e.Cancelled == false ? $"Network Progress: {0}% (finished)" : "Network Progress: {0}% (cancelled)";
TrainingProgressBar.Value = 0;
TotalEpochLabel.Text = $"Total Trained Epochs: {mnistNetwork.TotalEpochs.ToString("n2")}";
TestImage();
};
NetworkWorker.WorkerSupportsCancellation = true;
NumEpochsComboBox.SelectedIndex = 0;
}
public void CreateAndCompileModel(string jsonPath, string weightsDirectory)
{
Console.WriteLine("Creating model");
GlobalRandom.InitializeRandom();
model = new ConvolutionalNeuralNetwork(jsonPath);
Console.WriteLine("Model created");
model.Compile();
Console.WriteLine("Model compiled");
Console.WriteLine("Reading weights");
//ReadWeightsFromDirectory(weightsDirectory);
ReadWeightsFromDirectory(weightsDirectory);
}
public void CreateAndCompileModelMnist()
{
Console.WriteLine("Creating model");
GlobalRandom.InitializeRandom();
int imgSize = 28;
NoActivation noActivation = new NoActivation();
SoftmaxActivation softmaxActivation = new SoftmaxActivation();
model = new ConvolutionalNeuralNetwork(imgSize, "grayscale");
model.Add(new ConvolutionalLayer(8, 3, noActivation, "valid"));
model.Add(new MaxPoolingLayer());
model.Add(new FlattenLayer());
model.Add(new DenseLayer(10, softmaxActivation));
Console.WriteLine("Model created");
model.Compile();
Console.WriteLine("Model compiled");
}
public Trainer(ConvolutionalNeuralNetwork network, double learningRate)
{
LearningRate = learningRate;
_network = network;
}
