// Load an array of training set items from the csv file
public static List <TrainingItem> Load(string csvFile)
{
List <TrainingItem> trainingItems = new List <TrainingItem>();
string line;
string[] items;
// Read the file and display it line by line.
System.IO.StreamReader file = new System.IO.StreamReader(csvFile);
// Skip the first line since it is the header
file.ReadLine();
while ((line = file.ReadLine()) != null)
{
TrainingItem trainingItem = new TrainingItem();
items = line.Split(',');
trainingItem.FileName = items[0];
trainingItem.Label = items[1];
trainingItems.Add(trainingItem);
}
file.Close();
return(trainingItems);
}
// Load an array of training set items from the csv file
public static List<TrainingItem> Load(string csvFile)
{
List<TrainingItem> trainingItems = new List<TrainingItem>();
string line;
string[] items;
// Read the file and display it line by line.
System.IO.StreamReader file = new System.IO.StreamReader(csvFile);
// Skip the first line since it is the header
file.ReadLine();
while ((line = file.ReadLine()) != null)
{
TrainingItem trainingItem = new TrainingItem();
items = line.Split(',');
trainingItem.FileName = items[0];
trainingItem.Label = items[1];
trainingItems.Add(trainingItem);
}
file.Close();
return trainingItems;
}
// Converts the images in a directory into the appropriate dimension
// and saves them to the trianing set directory
public static void Convert()
{
// Load the images in the directory with training set images
DirectoryInfo di = new DirectoryInfo(@".\trainingset\tests");
FileInfo[] fi = di.GetFiles("*", SearchOption.TopDirectoryOnly);
// Empty the data csv file
System.IO.File.WriteAllText(trainingSetCSVFile, "FileName, Label\r\n");
foreach (FileInfo file in fi)
{
//// Convert the image into gray and the training width and height
//Image<Gray, Byte> image = new Image<Gray, Byte>(file.FullName);
//image = image.Resize(trainingImgWidth, trainingImgHeight, INTER.CV_INTER_CUBIC);
//// Save file in bmp
//string label = Path.GetFileNameWithoutExtension(file.FullName);
//string fileName = label + imageExt;
//image.Save(@".\trainingset\" + fileName);
List<RecognizerResult> list = Detect(file.FullName);
string fileName = Path.GetFileName(list[0].FileName);
File.Move(list[0].FileName, trainingSetDir + @"/" + fileName);
// Add item to CSV
TrainingItem item = new TrainingItem();
item.FileName = fileName;
item.Label = Path.GetFileNameWithoutExtension(list[0].FileName);
item.AddToCSV(trainingSetCSVFile);
}
}
private void btnSave_Click(object sender, EventArgs e)
{
int selectedIndex = lvResults.SelectedIndices[0];
if (selectedIndex >= 0 && selectedIndex < rcgnResultList.Count)
{
string sourceFile = rcgnResultList[selectedIndex].FileName;
string destFile = Program.trainingSetDir + @"\" + Path.GetFileName(sourceFile);
// Check if this image is already part of the training set
if (File.Exists(destFile))
{
System.Windows.Forms.MessageBox.Show("This image already exist in the training set.");
return;
}
// Copy the result image file to training set folder
System.IO.File.Copy(sourceFile, destFile, true);
// Add the item to the training set
TrainingItem item = new TrainingItem();
item.FileName = Path.GetFileName(sourceFile);
item.Label = txtLabel.Text;
item.AddToCSV(Program.trainingSetCSVFile);
// Process the update training set and update the display tabs
//ilTrainingSet.Images.Clear();
//lvTrainingSet.Items.Clear();
//ilEigenFaces.Images.Clear();
//lvEigenFaces.Items.Clear();
//Program.Train();
//showTrainingSet();
//showEigenFaces();
}
}
// Converts the images in a directory into the appropriate dimension
// and saves them to the trianing set directory
public static void Convert()
{
// Load the images in the directory with training set images
DirectoryInfo di = new DirectoryInfo(@".\trainingset\tests");
FileInfo[] fi = di.GetFiles("*", SearchOption.TopDirectoryOnly);
// Empty the data csv file
System.IO.File.WriteAllText(trainingSetCSVFile, "FileName, Label\r\n");
foreach (FileInfo file in fi)
{
//// Convert the image into gray and the training width and height
//Image<Gray, Byte> image = new Image<Gray, Byte>(file.FullName);
//image = image.Resize(trainingImgWidth, trainingImgHeight, INTER.CV_INTER_CUBIC);
//// Save file in bmp
//string label = Path.GetFileNameWithoutExtension(file.FullName);
//string fileName = label + imageExt;
//image.Save(@".\trainingset\" + fileName);
List <RecognizerResult> list = Detect(file.FullName);
string fileName = Path.GetFileName(list[0].FileName);
File.Move(list[0].FileName, trainingSetDir + @"/" + fileName);
// Add item to CSV
TrainingItem item = new TrainingItem();
item.FileName = fileName;
item.Label = Path.GetFileNameWithoutExtension(list[0].FileName);
item.AddToCSV(trainingSetCSVFile);
}
}
public static void Train()
{
List <TrainingItem> items = TrainingItem.Load(trainingSetCSVFile);
Image <Gray, Byte>[] trainingImages = new Image <Gray, Byte> [items.Count];
String[] labels = new String[items.Count];
for (int index = 0; index < items.Count; index++)
{
trainingImages[index] = new Image <Gray, byte>(trainingSetDir + @"\" + items[index].FileName);
labels[index] = Path.GetFileNameWithoutExtension(items[index].Label);
}
MCvTermCriteria termCrit = new MCvTermCriteria(maxEigenFaces, 0.001);
// Initialize Eigen Recognizer Object
recognizer = new EigenObjectRecognizer(
trainingImages,
labels,
eigenDistanceThreshold,
ref termCrit);
// Remove all old eigen images
DirectoryInfo eigenDir = new DirectoryInfo(eigenImageDir);
FileInfo[] eigenFiles = eigenDir.GetFiles();
foreach (FileInfo file in eigenFiles)
{
file.Delete();
}
// Save the average Image
recognizer.AverageImage.Save(averageImageName);
// Save all eigen images
for (int index = 0; index < recognizer.EigenImages.Length; index++)
{
String fileName = eigenImageName + (index + 1) + imageExt;
recognizer.EigenImages[index].ToBitmap().Save(fileName);
}
// Print eigen values
Console.WriteLine(String.Format("Num eigen values: {0}", recognizer.EigenValues.Length));
Console.WriteLine(String.Format("Num rows in eigen values: {0}", recognizer.EigenValues[0].Rows));
Console.WriteLine(String.Format("Num cols in eigen values: {0}", recognizer.EigenValues[0].Cols));
}
private void btnSave_Click(object sender, EventArgs e)
{
int selectedIndex = lvResults.SelectedIndices[0];
if (selectedIndex >= 0 && selectedIndex < rcgnResultList.Count)
{
string sourceFile = rcgnResultList[selectedIndex].FileName;
string destFile = Program.trainingSetDir + @"\" + Path.GetFileName(sourceFile);
// Check if this image is already part of the training set
if (File.Exists(destFile))
{
System.Windows.Forms.MessageBox.Show("This image already exist in the training set.");
return;
}
// Copy the result image file to training set folder
System.IO.File.Copy(sourceFile, destFile, true);
// Add the item to the training set
TrainingItem item = new TrainingItem();
item.FileName = Path.GetFileName(sourceFile);
item.Label = txtLabel.Text;
item.AddToCSV(Program.trainingSetCSVFile);
// Process the update training set and update the display tabs
//ilTrainingSet.Images.Clear();
//lvTrainingSet.Items.Clear();
//ilEigenFaces.Images.Clear();
//lvEigenFaces.Items.Clear();
//Program.Train();
//showTrainingSet();
//showEigenFaces();
}
}