public MainForm()
{
InitializeComponent();
// Список камер получаем
videoDevicesList = new FilterInfoCollection(FilterCategory.VideoInputDevice);
foreach (FilterInfo videoDevice in videoDevicesList)
{
cmbVideoSource.Items.Add(videoDevice.Name);
}
if (cmbVideoSource.Items.Count > 0)
{
cmbVideoSource.SelectedIndex = 0;
}
else
{
MessageBox.Show("А нет у вас камеры!", "Ошибочка", MessageBoxButtons.OK, MessageBoxIcon.Error);
}
controller = new Controller(new FormUpdateDelegate(UpdateFormFields));
// updateTmr = new System.Threading.Timer(Tick, evnt, 500, 100);
samplesSet = controller.processor.CreateSamplesSet();
Console.WriteLine("SAMPLE SET CREATED");
neuralNetwork.TrainOnDataSet(samplesSet, 20, 50);
}
public SamplesSet CreateSamplesSet()
{
var result = new SamplesSet();
ProcessClassSamples(result, Directory.GetFiles(@"..\..\Images\play"), FigureType.PLAY);
ProcessClassSamples(result, Directory.GetFiles(@"..\..\Images\stop"), FigureType.STOP);
ProcessClassSamples(result, Directory.GetFiles(@"..\..\Images\pause"), FigureType.PAUSE);
ProcessClassSamples(result, Directory.GetFiles(@"..\..\Images\forward"), FigureType.FORWARD);
ProcessClassSamples(result, Directory.GetFiles(@"..\..\Images\backward"), FigureType.BACKWARD);
return(result);
}
public override double TestOnDataSet(SamplesSet testSet)
{
double correct = 0.0;
foreach (Sample sample in testSet)
{
Run(sample);
sample.output = getOutput();
sample.processOutput();
if (sample.Correct())
{
correct += 1;
}
}
return(correct / testSet.Count);
}
private void ProcessClassSamples(SamplesSet samplesSet, string[] fileNames, FigureType type)
{
foreach (var fileName in fileNames)
{
// var image = getProcessedImage(new Bitmap(fileName));
// var newPath = fileName.Split('\\');
// newPath[newPath.Length - 1] = "processed_" + newPath.Last();
// image.Save(String.Join("\\", newPath));
var image = new Bitmap(fileName);
var inputs = new double[1000];
for (int i = 0; i < 500; i++)
{
inputs[i] = CountBlackPixels(GetBitmapColumn(image, i));
inputs[i + 500] = CountBlackPixels(GetBitmapRow(image, i));
}
samplesSet.AddSample(new Sample(inputs, 5, type));
Console.WriteLine($"Add sample {fileName}");
}
}
public override double TrainOnDataSet(SamplesSet samplesSet, int epochs_count, double acceptable_erorr,
bool parallel = true)
{
Random rnd = new Random();
var randomSamples = samplesSet.samples.OrderBy(x => rnd.Next());
var stopWatch = new Stopwatch();
stopWatch.Restart();
var guessLevel = 0.0;
for (int i = 1; i <= epochs_count; i++)
{
foreach (var sample in randomSamples)
{
if (Train((Sample)sample) == 0)
{
guessLevel += 1;
}
}
guessLevel /= samplesSet.samples.Count;
// updateDelegate((double)i / epochs_count, guessLevel, stopWatch.Elapsed);
if (1.0 - guessLevel < acceptable_erorr)
{
break;
}
}
Console.WriteLine($"Time elapsed: {stopWatch.Elapsed}");
stopWatch.Stop();
// updateDelegate(1.0, guessLevel, stopWatch.Elapsed);
return(guessLevel);
}
public abstract double TestOnDataSet(SamplesSet testSet);
public abstract double TrainOnDataSet(SamplesSet samplesSet, int epochs_count, double acceptable_erorr,
bool parallel = true);