private void AddToLog(string str)
{
LogBox.Text += str + Environment.NewLine;
LogBox.SelectionStart = LogBox.Text.Length;
LogBox.ScrollToCaret();
LogBox.Refresh();
}
/// <summary>
/// Adds a new line to the "status" window on the GUI
/// </summary>
/// <param name="message">The message to print</param>
/// <param name="newLine">Add a new line for the next message?</param>
public void UdpateStatus(string message, bool newLine)
{
// Add new line
if (newLine)
{
message = message + Environment.NewLine;
}
if (InvokeRequired)
{
// Invoke the logbox update
Invoke((MethodInvoker) delegate
{
LogBox.Text += message;
LogBox.Refresh();
});
}
else
{
LogBox.Text += message;
LogBox.Refresh();
}
}
private void AutoQueryButton_Click(object sender, EventArgs e)
{
double[] input = new double[784];
double[] answer = new double[10];
double[] output = new double[10];
string[] filePaths = Directory.GetFiles(@"MNIST Image\\", "*.bmp",
SearchOption.AllDirectories);
string image_addr;
string image_name;
Bitmap image;
Bitmap percent_graph = new Bitmap(percent_box.Width * 999, percent_box.Height);
double error_percent = 0;
LogBox.Clear();
using (Graphics g = Graphics.FromImage(percent_graph))
{
g.Clear(Color.Black);
}
for (int file = 0; file < filePaths.Length; file++)
{
error_percent = 0;
image_addr = filePaths[file];
image_name = Path.GetFileName(image_addr);
image = new Bitmap(image_addr);
LookWindow.Image = image;
LookWindow.Refresh();
for (int x = 0; x < answer.Length; x++)
{
if (network.neuron_arrays[network.neuron_arrays.Length - 1][x].name == "unnamed")
{
network.neuron_arrays[network.neuron_arrays.Length - 1][x].name = new DirectoryInfo(Directory.GetDirectories(@"MNIST Image")[x]).Name;
}
if (network.neuron_arrays[network.neuron_arrays.Length - 1][x].name == Path.GetFileName(Path.GetDirectoryName(image_addr)))
{
answer[x] = 1;
}
else
{
answer[x] = 0;
}
}
for (int y = 0; y < 28; y++)
{
for (int x = 0; x < 28; x++)
{
input[x + y * x] = 255 * (1 - Convert.ToDouble(image.GetPixel(x, y).GetBrightness()));
}
}
output = network.Query(input);
//LogBox.Clear();
double sum = 0;
bool mistake = false;
for (int x = 0; x < output.Length; x++)
{
if (answer[x] == 1)
{
error_percent += (1 - output[x]) / answer.Length;
}
else
{
error_percent += output[x] / answer.Length;
if (output[x] >= 0.5)
{
mistake = true;
}
}
sum += output[x];
}
if (mistake)
{
LogBox.Text += "Error in " + Path.GetFileName(filePaths[file]) + "\r\n";
}
LogBox.Refresh();
if (error_percent > 1)
{
error_percent = 1;
}
try
{
ErrorLabel.Text = Convert.ToString(error_percent * 100).Substring(0, 5) + "%";
}
catch (ArgumentOutOfRangeException)
{
ErrorLabel.Text = "100%";
}
ErrorLabel.Refresh();
using (Graphics g = Graphics.FromImage(percent_graph))
{
int a = Convert.ToInt32((percent_graph.Width / filePaths.Length) * file);
int b = Convert.ToInt32(percent_graph.Height - percent_graph.Height * error_percent);
int c = Convert.ToInt32(percent_graph.Width / filePaths.Length);
int d = Convert.ToInt32(percent_graph.Height * error_percent);
g.DrawRectangle(new Pen(Color.Green, 1), new Rectangle(a, b, c, d));
g.FillRectangle(new SolidBrush(Color.Green), new Rectangle(a, b, c, d));
}
percent_box.Image = percent_graph;
percent_box.Refresh();
}
}
private void MNIST_BMP_button_Click(object sender, EventArgs e)
{
OpenFileDialog openMNISTDialog = new OpenFileDialog();
string line;
openMNISTDialog.InitialDirectory = "\\";
openMNISTDialog.Filter = "All files (*.*)|*.*|MNIST files(*.csv)|*.csv";
openMNISTDialog.FilterIndex = 2;
openMNISTDialog.RestoreDirectory = true;
if (openMNISTDialog.ShowDialog() == DialogResult.OK)
{
string filename = openMNISTDialog.FileName;
string output_addr = "";
var fileStream = new FileStream(filename, FileMode.Open, FileAccess.Read);
using (var streamReader = new StreamReader(fileStream, Encoding.UTF8))
{
while (true)
{
Bitmap Image = new Bitmap(28, 28);
char tmp;
string number = "";
int x = 0;
int y = 0;
int word = 0;
int brightness;
output_addr = "";
line = streamReader.ReadLine();
if (line == null)
{
break;
}
//textBox1.Text = Convert.ToString(line.Length);
for (int symbol = 0; symbol < line.Length; symbol++)
{
tmp = Convert.ToChar(line.Substring(symbol, 1));
if (tmp == ',' && word == 0)
{
//textBox1.Text = Convert.ToString(tmp);
output_addr += line.Substring(0, symbol);
number = "";
word++;
}
else if (tmp == ',')
{
brightness = Convert.ToInt32(number);
number = "";
Image.SetPixel(x, y, Color.FromArgb(255, 255 - brightness, 255 - brightness, 255 - brightness));
x++;
if (x == 28)
{
x = 0;
y++;
}
}
else
{
number = number + tmp;
if (symbol == line.Length - 1)
{
brightness = Convert.ToInt32(number);
Image.SetPixel(x, y, Color.FromArgb(255, 255 - brightness, 255 - brightness, 255 - brightness));
}
}
//Image.SetPixel(x, y, Color.FromArgb(152, 15, 198, 255));
}
LookWindow.Image = Image;
LookWindow.Refresh();
try
{
output_addr = @"MNIST Image\\" + output_addr + "\\" + output_addr + "_[" + Convert.ToString(Directory.GetFiles(@"MNIST Image\\" + output_addr).Length) + "].bmp";
}
catch (System.IO.DirectoryNotFoundException)
{
Directory.CreateDirectory(@"MNIST Image\\" + output_addr);
output_addr = @"MNIST Image\\" + output_addr + "\\" + output_addr + "_[0].bmp";
}
LogBox.Text += Convert.ToString(output_addr) + "\r\n";
LogBox.Refresh();
Image.Save(output_addr);
}
}
}
}
private void TrainButton_Click(object sender, EventArgs e)
{
int epochs = 100;
double index = 0.3;
string[] filePaths = Directory.GetFiles(@"MNIST Image\\", "*.bmp",
SearchOption.AllDirectories);
string image_addr;
string image_name;
Bitmap image;
Bitmap percent_graph = new Bitmap(percent_box.Width * 999, percent_box.Height);
double error_percent = 0;
Random rand = new Random();
try
{
epochs = Convert.ToInt32(EpochBox.Text);
}
catch (FormatException)
{
epochs = 100;
EpochBox.Text = "100";
}
try
{
index = Convert.ToInt32(IndexBox.Text);
}
catch (FormatException)
{
index = 0.3;
IndexBox.Text = "0,3";
}
if (EpochBox.Text == "")
{
EpochBox.Text = "100";
}
if (IndexBox.Text == "")
{
IndexBox.Text = "0,3";
}
using (Graphics g = Graphics.FromImage(percent_graph))
{
g.Clear(Color.Black);
}
for (int epoch = 0; epoch < epochs; epoch++)
{
error_percent = 0;
for (int file = 0; file < filePaths.Length; file++)
{
image_addr = filePaths[rand.Next(0, filePaths.Length)];
image_name = Path.GetFileName(image_addr);
image = new Bitmap(image_addr);
LookWindow.Image = image;
LookWindow.Refresh();
for (int x = 0; x < answer.Length; x++)
{
if (network.neuron_arrays[network.neuron_arrays.Length - 1][x].name == "unnamed")
{
network.neuron_arrays[network.neuron_arrays.Length - 1][x].name = new DirectoryInfo(Directory.GetDirectories(@"MNIST Image")[x]).Name;
}
if (network.neuron_arrays[network.neuron_arrays.Length - 1][x].name == Path.GetFileName(Path.GetDirectoryName(image_addr)))
{
answer[x] = 1;
}
else
{
answer[x] = 0;
}
}
for (int y = 0; y < 28; y++)
{
for (int x = 0; x < 28; x++)
{
input[x + y * x] = 1 - Convert.ToDouble(image.GetPixel(x, y).GetBrightness());
}
}
output = network.Query(input);
LogBox.Clear();
double sum = 0;
for (int x = 0; x < output.Length; x++)
{
if (answer[x] == 1)
{
error_percent += (1 - output[x]) / filePaths.Length;
}
else
{
error_percent += output[x] / filePaths.Length;
}
sum += output[x];
}
for (int x = 0; x < output.Length; x++)
{
LogBox.Text += network.neuron_arrays[network.neuron_arrays.Length - 1][x].name + " => " + Convert.ToString((output[x] / sum) * 100).Substring(0, 4) + "%\r\n";
}
LogBox.Refresh();
network.Train(input, answer, index);
//Thread.Sleep(500);}
}
if (error_percent > 1)
{
error_percent = 1;
}
try
{
ErrorLabel.Text = Convert.ToString(error_percent * 100).Substring(0, 5) + "%";
}
catch (ArgumentOutOfRangeException)
{
ErrorLabel.Text = "100%";
}
ErrorLabel.Refresh();
using (Graphics g = Graphics.FromImage(percent_graph))
{
int a = Convert.ToInt32((percent_graph.Width / epochs) * epoch);
int b = Convert.ToInt32(percent_graph.Height - percent_graph.Height * error_percent);
int c = Convert.ToInt32(percent_graph.Width / epochs);
int d = Convert.ToInt32(percent_graph.Height * error_percent);
g.DrawRectangle(new Pen(Color.Green, 1), new Rectangle(a, b, c, d));
g.FillRectangle(new SolidBrush(Color.Green), new Rectangle(a, b, c, d));
}
percent_box.Image = percent_graph;
percent_box.Refresh();
}
/*ouble[] input = {1.0, 0.5, -1.5};
* double[] answer = {0.20813168, 0.22549015, 0.99742175};
* double index = 0.3;
* double[] output = new double[3];
* output = network.Query(input);
* network.Train(input, answer, index);
* output = network.Query(input);*/
}
private void Form1_ResizeEnd(object sender, EventArgs e)
{
LogBox.Refresh();
}