public Load_L2BOF(int classnum, string src)
{
SecondLayerDic = null;
directory = src;
classNum = classnum;
bowTrainer = new BOWKMeansTrainer(classNum, new MCvTermCriteria(10, 0.01), 3,
KMeansInitType.PPCenters);
IFormatter formatter = new BinaryFormatter();
FileStream fs = File.OpenRead(directory + "\\obj\\dic.xml");
SecondLayerDic = (Matrix<float>)formatter.Deserialize(fs);
fs.Dispose();
bowDe.SetVocabulary(SecondLayerDic);
SecondLayerSVM = new SVM();
SecondLayerSVM.Load(directory + "\\obj\\svm.xml");
Console.WriteLine("Finished Loading L2 SVM.");loaded = true;
}
public Load_L1BOF(int classnum,string loc)
{
prefix = loc;
topLayerDic = null;
classNum = classnum;
bowTrainer = new BOWKMeansTrainer(classNum, new MCvTermCriteria(10, 0.01), 3,
KMeansInitType.PPCenters);
IFormatter formatter = new BinaryFormatter();
FileStream fs = File.OpenRead(loc + "obj\\dic.xml");
topLayerDic = (Matrix<float>)formatter.Deserialize(fs);
fs.Dispose();
bowDe.SetVocabulary(topLayerDic);
topLayerSVM = new SVM();
topLayerSVM.Load(loc + "obj\\svm.xml");
Console.WriteLine("Finished Loading L1 SVM.");
loaded = true;
}
private void button7_Click(object sender, EventArgs e)
{
///////////////////////////////////////////// Test ///////////////////////////////////////
//Create a matrix
//Matrix<Byte> matrix1 = new Matrix<Byte>(10, 10);
//Byte element = 0;
////Set the elements
//for (int i = 0; i < 10; i++)
//{
// for (int j = 0; j < 10; j++)
// {
// matrix1.Data[i, j] = 1;
// //matrix1.Data[i, j] = element;
// element++;
// }
//}
//int WW = matrix1.Cols;
//int HH = matrix1.Rows;
///////////////////////////////////////////// Test ///////////////////////////////////////
int W = Train_I.Cols;
int H = Train_I.Rows;
int rW = (mW - 1) / 2;
int rH = (mH - 1) / 2;
double pp = 0;
double ee = 0;
int ind = 0;
int count;
int maskCenter_H_start = rH + 1;
int maskCenter_W_start = rW + 1;
int maskCenter_H_end = H - rH;
int maskCenter_W_end = W - rW;
double[] fVave = new double[ ( maskCenter_H_end - maskCenter_H_start) * (maskCenter_W_end - maskCenter_W_start) ];
double[] fVstd = new double[( maskCenter_H_end - maskCenter_H_start) * (maskCenter_W_end - maskCenter_W_start) ];
double[] fVentropy = new double[( maskCenter_H_end - maskCenter_H_start) * (maskCenter_W_end - maskCenter_W_start) ];
classV = new Matrix<float>( (maskCenter_H_end - maskCenter_H_start) * (maskCenter_W_end - maskCenter_W_start), 1);
// Mask scan over the whole image "gR_R_Picked_Map"
for (int i = maskCenter_H_start; i < maskCenter_H_end; i++)
{
for (int j = maskCenter_W_start; j < maskCenter_W_end; j++)
{
double sum = 0;
double ave = 0;
double sumDiff = 0;
double std = 0;
entropy = 0;
count = 0;
// initial value all 0, since every mask should count independently
for (int k = 0; k < mH; k++)
{
for (int l = 0; l < mW; l++)
{
int mask_x = j - rW + l - 1;
int mask_y = i - rH + k - 1;
sum = sum + gR_R_Picked_Map[ mask_y, mask_x ]; // Bug????
//sum = sum + matrix1[i - rH + k - 1 , j - rW + l - 1 ]; // test
// judge mask is PART of TREE or NOT
if (Class_I.Data[ mask_y, mask_x , 0] == 1)
{
count ++;
}
}
} // end of mask
// To detect the binary values in the mask and record it in classV
float maskSampleCount = mH * mW;
float mask_plant_percent = (count /maskSampleCount);
if ( mask_plant_percent > 0.7 )
{
classV[ind, 0] = 1;
}
else
{
classV[ind, 0] = 0;
}
// Ave
ave = sum / (maskSampleCount);
// StandardDeviation
for (int k = 0; k < mH; k++)
{
for (int l = 0; l < mW; l++)
{
int mask_x = j - rW + l - 1;
int mask_y = i - rH + k - 1;
sumDiff = sumDiff + Math.Pow((gR_R_Picked_Map[ mask_y, mask_x] - ave), 2);
// sumDiff = sumDiff + Math.Pow((matrix1[i - rH + k - 1, j - rW + l - 1] - ave), 2); //test
}
}
std = Math.Sqrt(sumDiff / (maskSampleCount - 1));
// Entropy
for (int k = 1; k < mH; k++)
{
for (int l = 1; l < mW; l++)
{
int mask_x = j - rW + l - 1;
int mask_y = i - rH + k - 1;
pp = gR_R_Picked_Map[mask_y, mask_x] / sum;
//p = matrix1[i - rH + k - 1, j - rW + l - 1] / sum;
if (pp > 0) // Log p, p > 0
{
ee = Math.Log(pp);
}
else
{
ee = 0;
}
//System.Console.WriteLine("-");
//System.Console.WriteLine(p);
//System.Console.WriteLine(ee);
//System.Console.WriteLine("-");
entropy = entropy + (-(pp * ee));
}
}
fVave[ind] = ave;
fVstd[ind] = std;
fVentropy[ind] = entropy;
ind++;
} // for i
} // for j
trainData_all = new Matrix<float>(ind, 3);
for (int i = 0; i < ind; i++)
{
trainData_all.Data[i, 0] = (float)fVave[i];
trainData_all.Data[i, 1] = (float)fVstd[i];
trainData_all.Data[i, 2] = (float)fVentropy[i];
}
using (Emgu.CV.ML.SVM model = new Emgu.CV.ML.SVM())
{
SVMParams p = new SVMParams();
/* p.KernelType = Emgu.CV.ML.MlEnum.SVM_KERNEL_TYPE.LINEAR;
p.SVMType = Emgu.CV.ML.MlEnum.SVM_TYPE.C_SVC;
p.C = 1;
p.TermCrit = new MCvTermCriteria(100, 0.00001);*/
p.KernelType = Emgu.CV.ML.MlEnum.SVM_KERNEL_TYPE.RBF;
p.SVMType = Emgu.CV.ML.MlEnum.SVM_TYPE.C_SVC;
p.Gamma = 0.1;
p.Coef0 = 0;
p.Degree = 2;
p.C = 1;
p.TermCrit = new MCvTermCriteria(500, 0.00001); //迭代的終止準則,最大迭代次數500次,結果的精確性
//bool trained = model.Train(trainData, trainClasses, null, null, p);
bool trained = model.TrainAuto(trainData_all, classV, null, null, p.MCvSVMParams, 5);
model.Save(@"D:\Code\Data\train_func.xml");
}
System.Console.WriteLine(" END OF button7_Click Masking ");
///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
Image<Bgr, byte> predict_image = Train_I2.Clone();
using (Emgu.CV.ML.SVM model = new Emgu.CV.ML.SVM())
{
model.Load(@"D:\Code\Data\train_func.xml");
int ind_predict = 0;
for (int i = maskCenter_H_start; i < maskCenter_H_end; i++)
for (int j = maskCenter_W_start; j < maskCenter_W_end; j++)
{
Matrix<float> feature_test = TakeRow(trainData_all.Data, ind_predict);
//Matrix<float> feature_test = trainData_all.GetRow( ind_predict ) ;
// System.Console.WriteLine(feature_test);
float predict_class = model.Predict( feature_test );
if (predict_class == 1)
{
for (int k = 0; k < mH; k++)
for (int l = 0; l < mW; l++)
{
int mask_x = j - rW + l -1;
int mask_y = i - rH + k -1;
predict_image.Data[ mask_y, mask_x , 0] = 255;
predict_image.Data[mask_y , mask_x , 2] = 255;
}
}
ind_predict++;
// System.Console.WriteLine(predict_class);
} // end of all
//String name2 = name.Insert(name.Length - 4, "_predict");
String name3 = "predict";
predict_image.Save(@"D:\Code\Data\predict.bmp");
pictureBox9.Image = Image.FromFile(@"D:\Code\Data\predict.bmp");
// Image<Bgr, byte> predict_image = Train_I2.Clone();
System.Console.WriteLine(" END OF Predict ");
}
}
public static SVM CreateSvm()
{
var model = new SVM();
model.Load(ModelFileName);
return model;
}
private void button1_Click(object sender, EventArgs e)
{
folderBrowserDialog1.SelectedPath = @"D:\Code\Data";
try
{
if (folderBrowserDialog1.ShowDialog() == DialogResult.OK)
{
String[] files = System.IO.Directory.GetFiles(folderBrowserDialog1.SelectedPath);
foreach (String name in files)
{
if (name.Contains("_class") )/////////有分類標籤的影像
{
/* if (Class_I != null)
{
Class_I.Dispose(); Class_I = null;
}
if (Train_I != null)
{
Train_I.Dispose(); Train_I = null;
}
//////////
String name2 = name.Replace("_class", "");
name2 = name2.Replace("bmp", "tif");
if (files.Contains(name2))
{
Class_I = new Image<Bgr, Byte>(name); //////類別
Train_I = new Image<Bgr, Byte>(name2); //////原影像
if (trainData_all != null)
{
trainData_all.Dispose(); trainData_all = null;
}
Cal_features();
Image<Bgr, byte> predict_image = Train_I.Clone();
using (Emgu.CV.ML.SVM model = new Emgu.CV.ML.SVM())
{
model.Load(@"D:\程式區\Plant_training\train_func.xml");
for (int i = 0; i < Train_I.Height; i++)
for (int j = 0; j < Train_I.Width; j++)
{
float predict_class = model.Predict(trainData_all.GetRow(i * Train_I.Width + j));
if (predict_class == 1)
{
predict_image.Data[i, j, 0] = 255;
predict_image.Data[i, j, 2] = 255;
}
}
}
predict_image.Save(@"D:\程式區\Plant_training\predict.tif");
}*/
}
else
{
if (Train_I != null)
{
Train_I.Dispose(); Train_I = null;
}
if (Class_I != null)
{
Class_I.Dispose(); Class_I = null;
}
Train_I = new Image<Bgr, Byte>(name).Resize(scale, Emgu.CV.CvEnum.INTER.CV_INTER_LINEAR); //////原影像
if (trainData_all != null)
{
trainData_all.Dispose(); trainData_all = null;
}
Cal_features();
Image<Bgr, byte> predict_image = Train_I.Clone();
using (Emgu.CV.ML.SVM model = new Emgu.CV.ML.SVM())
{
model.Load(@"D:\Code\Data\train_func.xml");
for (int i = 0; i < Train_I.Height; i++)
for (int j = 0; j < Train_I.Width; j++)
{
Matrix<float> feature_test = trainData_all.GetRow(i * Train_I.Width + j);
int aaaaa = 1;
float predict_class = model.Predict(feature_test);
if (predict_class == 1)
{
predict_image.Data[i, j, 0] = 255;
predict_image.Data[i, j, 2] = 255;
}
}
}
String name2 = name.Insert(name.Length - 4, "_predict");
//String name2 = "predict";
predict_image.Save(name2);
}
}
}
}
catch
{
MessageBox.Show("讀取錯誤!");
}
}
//SVMParams _parameters
public SVMClassifier(string filename)
{
model = new SVM();
model.Load(filename);
}
/*
public void OnContactRecordGesture(object sender, FrameReceivedEventArgs e)
{
if (isTouching)
{
if (normalizedImage == null)
{
e.TryGetRawImage(
ImageType.Normalized,
0, 0,
InteractiveSurface.DefaultInteractiveSurface.Width,
InteractiveSurface.DefaultInteractiveSurface.Height,
out normalizedImage,
out normalizedMetrics);
}
else //updates raw image data
{
e.UpdateRawImage(
ImageType.Normalized,
normalizedImage,
0, 0,
InteractiveSurface.DefaultInteractiveSurface.Width,
InteractiveSurface.DefaultInteractiveSurface.Height);
}
capture.OnContactRecordHelper(normalizedImage, normalizedMetrics);
}
}
*/
/// <summary>
/// Controls what occurs when the user stops a gesture
/// NOTE: This event fires more than it ought. So we have it ignore any data shorter than 600 millisecondss
/// </summary>
public void OffContactStopRecord(object sender, ContactEventArgs e)
{
capture.OffContactHelper();
if (capture.totalMillisec > 600)
{
lock (obj)
{
writeToCSV();
}
capture.calculateRatio(touchManager);
advanceGesture();
Gesture gnew = stListToGesture(touchManager);
float result;
using (SVM svm = new SVM())
{
svm.Load(@"C:\Users\faculty\Desktop\svm-function3coord16.xml");
result = svm.Predict(getProcessedGesture(gnew));
}
if (result == 1)
{
gestureOwner = "Evan";
}
else
{
gestureOwner = "Stevie";
}
}
}