private void MnistDemo()
{
Directory.CreateDirectory(mnistFolder);
string trainingLabelFilePath = Path.Combine(mnistFolder, trainingLabelFile);
string trainingImageFilePath = Path.Combine(mnistFolder, trainingImageFile);
string testingLabelFilePath = Path.Combine(mnistFolder, testingLabelFile);
string testingImageFilePath = Path.Combine(mnistFolder, testingImageFile);
// Download Mnist files if needed
Console.WriteLine("Downloading Mnist training files...");
DownloadFile(urlMnist + trainingLabelFile, trainingLabelFilePath);
DownloadFile(urlMnist + trainingImageFile, trainingImageFilePath);
Console.WriteLine("Downloading Mnist testing files...");
DownloadFile(urlMnist + testingLabelFile, testingLabelFilePath);
DownloadFile(urlMnist + testingImageFile, testingImageFilePath);
// Load data
Console.WriteLine("Loading the datasets...");
this.training = MnistReader.Load(trainingLabelFilePath, trainingImageFilePath);
this.testing = MnistReader.Load(testingLabelFilePath, testingImageFilePath);
if (this.training.Count == 0 || this.testing.Count == 0)
{
Console.WriteLine("Missing Mnist training/testing files.");
Console.ReadKey();
return;
}
// Create network
this.net = FluentNet.Create(24, 24, 1)
.Conv(5, 5, 8).Stride(1).Pad(2)
.Relu()
.Pool(2, 2).Stride(2)
.Conv(5, 5, 16).Stride(1).Pad(2)
.Relu()
.Pool(3, 3).Stride(3)
.FullyConn(10)
.Softmax(10)
.Build();
this.trainer = new AdadeltaTrainer(this.net)
{
BatchSize = 20,
L2Decay = 0.001,
};
Console.WriteLine("Convolutional neural network learning...[Press any key to stop]");
do
{
var sample = this.SampleTrainingInstance();
this.Step(sample);
} while (!Console.KeyAvailable);
}
private void TrainNetworkForTactile(double availableLoss)
{
trainer = new AdadeltaTrainer(net)
{
// Количество обрабатываемых образцов за заход
BatchSize = 15,
// Регуляризация - штраф на наибольший вес
L2Decay = 0.001,
};
do
{
var sample = PrepareTrainingSample();
TrainingStep(sample);
} while (loss > availableLoss);
}
private void MnistDemo()
{
Directory.CreateDirectory(mnistFolder);
string trainingLabelFilePath = Path.Combine(mnistFolder, trainingLabelFile);
string trainingImageFilePath = Path.Combine(mnistFolder, trainingImageFile);
string testingLabelFilePath = Path.Combine(mnistFolder, testingLabelFile);
string testingImageFilePath = Path.Combine(mnistFolder, testingImageFile);
// Download Mnist files if needed
Console.WriteLine("Downloading Mnist training files...");
DownloadFile(urlMnist + trainingLabelFile, trainingLabelFilePath);
DownloadFile(urlMnist + trainingImageFile, trainingImageFilePath);
Console.WriteLine("Downloading Mnist testing files...");
DownloadFile(urlMnist + testingLabelFile, testingLabelFilePath);
DownloadFile(urlMnist + testingImageFile, testingImageFilePath);
// Load data
Console.WriteLine("Loading the datasets...");
this.training = MnistReader.Load(trainingLabelFilePath, trainingImageFilePath);
this.testing = MnistReader.Load(testingLabelFilePath, testingImageFilePath);
if (this.training.Count == 0 || this.testing.Count == 0)
{
Console.WriteLine("Missing Mnist training/testing files.");
Console.ReadKey();
return;
}
// Create network
this.net = new Net();
this.net.AddLayer(new InputLayer(24, 24, 1));
this.net.AddLayer(new ConvLayer(5, 5, 8)
{
Stride = 1, Pad = 2
});
this.net.AddLayer(new ReluLayer());
this.net.AddLayer(new PoolLayer(2, 2)
{
Stride = 2
});
this.net.AddLayer(new ConvLayer(5, 5, 16)
{
Stride = 1, Pad = 2
});
this.net.AddLayer(new ReluLayer());
this.net.AddLayer(new PoolLayer(3, 3)
{
Stride = 3
});
this.net.AddLayer(new FullyConnLayer(10));
this.net.AddLayer(new SoftmaxLayer(10));
this.trainer = new AdadeltaTrainer(this.net)
{
BatchSize = 20,
L2Decay = 0.001,
};
Console.WriteLine("Convolutional neural network learning...[Press any key to stop]");
do
{
var sample = this.SampleTrainingInstance();
this.Step(sample);
} while (!Console.KeyAvailable);
}
private void MnistDemo()
{
Directory.CreateDirectory(mnistFolder);
string trainingLabelFilePath = Path.Combine(mnistFolder, trainingLabelFile);
string trainingImageFilePath = Path.Combine(mnistFolder, trainingImageFile);
string testingLabelFilePath = Path.Combine(mnistFolder, testingLabelFile);
string testingImageFilePath = Path.Combine(mnistFolder, testingImageFile);
// Download Mnist files if needed
Console.WriteLine("Downloading Mnist training files...");
DownloadFile(urlMnist + trainingLabelFile, trainingLabelFilePath);
DownloadFile(urlMnist + trainingImageFile, trainingImageFilePath);
Console.WriteLine("Downloading Mnist testing files...");
DownloadFile(urlMnist + testingLabelFile, testingLabelFilePath);
DownloadFile(urlMnist + testingImageFile, testingImageFilePath);
// Load data
Console.WriteLine("Loading the datasets...");
this.training = MnistReader.Load(trainingLabelFilePath, trainingImageFilePath).Where(p => numbers.Contains(p.Label)).ToList();
this.testing = MnistReader.Load(testingLabelFilePath, testingImageFilePath).Where(p => numbers.Contains(p.Label)).ToList();
if (this.training.Count == 0 || this.testing.Count == 0)
{
Console.WriteLine("Missing Mnist training/testing files.");
Console.ReadKey();
return;
}
Console.WriteLine($"datasets training: {this.training.Count}, testing: {this.testing.Count}");
//ExtractImages();
var netFile = Path.Combine(mnistFolder, $"net{string.Join("", numbers)}.bin");
if (File.Exists(netFile))
{
Console.WriteLine($"load {netFile}?");
if (Console.ReadKey(true).Key == ConsoleKey.Enter)
{
Console.WriteLine($"loading...");
this.net = Net.Load(netFile);
}
}
if (this.net == null)
{
this.net = new Net();
this.net.AddLayer(new InputLayer(24, 24, 1));
this.net.AddLayer(new ConvLayer(5, 5, 8)
{
Stride = 1, Pad = 2
});
this.net.AddLayer(new ReluLayer());
//this.net.AddLayer(new PoolLayer(2, 2) { Stride = 2 });
this.net.AddLayer(new ConvLayer(5, 5, 16)
{
Stride = 1, Pad = 2
});
this.net.AddLayer(new ReluLayer());
//this.net.AddLayer(new PoolLayer(3, 3) { Stride = 3 });
this.net.AddLayer(new FullyConnLayer(numbers.Length));
//this.net.AddLayer(new DropOutLayer());
this.net.AddLayer(new SoftmaxLayer(numbers.Length));
}
this.trainer = new AdadeltaTrainer(this.net)
{
BatchSize = 20,
L2Decay = 0.001,
};
Console.WriteLine($"limit cpu cores to 1?");
if (Console.ReadKey(true).Key == ConsoleKey.Enter)
{
using (Process Proc = Process.GetCurrentProcess())
{
long AffinityMask = (long)Proc.ProcessorAffinity;
AffinityMask = 1;
Proc.ProcessorAffinity = (IntPtr)AffinityMask;
}
}
Console.WriteLine("Training...[Press any key to stop]");
bool ok = false;
while (!ok)
{
do
{
var sample = this.SampleTrainingInstance();
ok = this.Step(sample);
}while (!ok && !Console.KeyAvailable);
if (!ok)
{
Console.ReadKey(true);
}
Console.WriteLine($"stop? [ENTER: continue]");
ok = !(Console.ReadKey(true).Key == ConsoleKey.Enter);
}
Console.WriteLine($"save {netFile}?");
if (Console.ReadKey(true).Key == ConsoleKey.Enter)
{
var f = Path.Combine(mnistFolder, netFile);
Console.WriteLine($"saving...");
net.Save(f);
}
Console.WriteLine("done.");
Console.ReadKey();
}
private const string folder = @"..\..\..\char\"; // para que saia da pasta 'debug' encontre os dados de treino
private void MnistTrain()
{
// carrega os chars
Console.WriteLine("carregando datasets...");
if (File.Exists("dataset_training.dat") || File.Exists("dataset_testing.dat"))
{ // se já tiver os datasets tratados, carrega eles
training = useful.ReadObject <List <MnistEntry> >(File.ReadAllBytes("dataset_training.dat"));
testing = useful.ReadObject <List <MnistEntry> >(File.ReadAllBytes("dataset_testing.dat"));
}
else
{ // caso contrário, carrega a partir das imagens
training = MnistReader.Load(folder, true);
testing = MnistReader.Load(folder, false);
File.WriteAllBytes("dataset_training.dat", useful.WriteObject <List <MnistEntry> >(training));
File.WriteAllBytes("dataset_testing.dat", useful.WriteObject <List <MnistEntry> >(testing));
}
Random rnd = new Random();
training = training.OrderBy(x => rnd.Next()).ToList();
testing = testing.OrderBy(x => rnd.Next()).ToList();
if (training.Count == 0 || testing.Count == 0)
{
Console.WriteLine("ajuste o diretório dos arquivos de treino/teste.");
Console.ReadKey();
return;
}
// cria uma CNN simples
net = new Net();
net.AddLayer(new InputLayer(24, 24, 1)); //tamanho que eu escalei as imagens
net.AddLayer(new ConvLayer(5, 5, 20)
{
Stride = 1, Pad = 2, Activation = Activation.Relu
});
net.AddLayer(new PoolLayer(2, 2)
{
Stride = 2
});
net.AddLayer(new ConvLayer(5, 5, 35)
{
Stride = 1, Pad = 2, Activation = Activation.Relu
});
net.AddLayer(new PoolLayer(3, 3)
{
Stride = 3
});
net.AddLayer(new FullyConnLayer(50));
net.AddLayer(new SoftmaxLayer(33));
// este é o meu otimizador
// ver depois http://int8.io/comparison-of-optimization-techniques-stochastic-gradient-descent-momentum-adagrad-and-adadelta/#AdaDelta
trainer = new AdadeltaTrainer(net)
{
BatchSize = 20,
L2Decay = 0.001,
};
Console.WriteLine("CNN treinando... aperte uma tecla para parar");
// o passo
do
{
var sample = SampleTrainingInstance();
// // só pra ver no matlab a imagem (no matlab eu uso o script imag.m)
//if (File.Exists("oi.txt")) { File.Delete("oi.txt"); }
//for (int i = 0; i < sample.Volume.Weights.Length; i++)
// using (StreamWriter file = new StreamWriter(@"oi.txt", true))
// file.WriteLine(((sample.Volume.Weights[i] * 255.0)).ToString());
Step(sample);
} while (!Console.KeyAvailable);
}