private static void Main1()
{
var trainPath = @"C:\Programming\ConvolutionNeuralNetwork\Images\Train";
var testPath = @"C:\Programming\ConvolutionNeuralNetwork\Images\Test";
var output = new List <string> {
"Zero", "One", "Two", "Three", "Four", "Five", "Six", "Seven", "Eight", "Nine"
};
var network = new ConvolutionNN
{
InputLayer = new InputLayer3D(28, 28),
OutputLayer = new OutputLayer(10),
DataProvider = new ClassFolderImageProvider(trainPath, testPath, new Size(28, 28), output),
Classes = output
};
network.HiddenLayers.Add(new ConvolutionalLayer(3, 2, 2, 3)
{
AutoPadding = true
});
network.HiddenLayers.Add(new PoolingLayer());
network.HiddenLayers.Add(new ConvolutionalLayer(3, 2, 2, 3)
{
AutoPadding = true
});
network.HiddenLayers.Add(new PoolingLayer());
network.HiddenLayers.Add(new NarrowLayer());
network.HiddenLayers.Add(new PerceptronLayer(100, 147));
network.HiddenLayers.Add(new PerceptronLayer(50, 100));
network.HiddenLayers.Add(new PerceptronLayer(10, 50));
}
private static void OneData()
{
var trainImage = @"C:\Programming\ConvolutionNeuralNetwork\Images\Train\One\One_0.jpg";
var expected = new Array3D(0, 1, 0, 0, 0, 0, 0, 0, 0, 0);
var network = new ConvolutionNN
{
InputLayer = new InputLayer3D(28, 28),
OutputLayer = new OutputLayer(10),
DataProvider = new OneIP(trainImage, expected)
};
network.HiddenLayers.Add(new ConvolutionalLayer(3, 2, 2, 3));
network.HiddenLayers.Add(new PoolingLayer());
network.HiddenLayers.Add(new ConvolutionalLayer(3, 2, 2, 3));
network.HiddenLayers.Add(new PoolingLayer());
network.HiddenLayers.Add(new NarrowLayer());
network.HiddenLayers.Add(new PerceptronLayer(100, 147));
network.HiddenLayers.Add(new PerceptronLayer(50, 100));
network.HiddenLayers.Add(new PerceptronLayer(10, 50));
var trainer = new FCTrainer(network, 10, 1, network.DataProvider);
trainer.Train(1000);
}
