void mitk::ModelSignalImageGenerator::SortParameterImages()
{
ParameterVectorType inputImages(this->m_ParameterInputMap.size());
unsigned int i = 0;
for (ParameterMapType::const_iterator pos = m_ParameterInputMap.begin(); pos != m_ParameterInputMap.end(); ++pos)
{
i = pos->first;
inputImages[i] = pos->second;
}
this->m_InputParameterImages = inputImages;
}
// ------------------------------------------------------------------------------------------
// Classify STL Test Image Set and save each file separated class into Image/class no. folder
//
void ClassifySTL()
{
CImageObject debugImage;
//debugImage.CreateReshape(serialized, rt.Width(), rt.Height(), 3);
//m_ImageDebugDlg.DrawImage(debugImage);
#if 1
CMatLoader inputImages("trainImages2000");
CMatLoader inputLabels("trainLabels2000");
CMatLoader pooledFeatures("pooledFeaturesTrain2000");
CMatLoader softmaxOptTheta("softmaxOptTheta2000");
#else
CMatLoader inputImages("testImages");
CMatLoader inputLabels("testLabels");
// 3600 x m
CMatLoader pooledFeatures("pooledFeaturesTest");
CMatLoader softmaxOptTheta("softmaxOptTheta2000");
#endif
// 4 x 3600
vector <BYTE *> vecImages;
SerializeSTLImage(inputImages, vecImages);
// CMatLoader testImages("testImages");
CStopWatch w;
vector <short> vecLabel;
vector <float> vecConfidence;
ClassifySoftmaxRegression(pooledFeatures, softmaxOptTheta, vecLabel, vecConfidence);
int count = 0;
int i;
//CLog log("result2.csv", true);
for (i=0;i<(int)vecLabel.size();i++)
{
if (vecLabel[i] == (int)inputLabels.data[i])
count++;
//log.WriteLog("%d, %d, %d\n", vecLabel[i] , (int)inputLabels.data[i], vecLabel[i] - (int)inputLabels.data[i]);
}
printf("Classification finished accuracy : %f %%\n", (float)count / (float)vecLabel.size() * 100.0f);
printf("Elapsed time %.0f msec\n", w.CheckTime());
mkdir("ResultImage");
mkdir("ResultImage\\1");
mkdir("ResultImage\\2");
mkdir("ResultImage\\3");
mkdir("ResultImage\\4");
mkdir("ResultImage\\5");
if (1)
{
for (i=0;i<(int)vecImages.size();i++)
//for (i=0;i<100;i++)
{
debugImage.CreateReshape(vecImages[i], 64, 64, 3);
CString str;
str.Format("ResultImage\\%d\\%d.bmp", vecLabel[i],i);
debugImage.SaveToBMP(str.GetBuffer(0));
}
// Classification(pooledFeaturesTest, softmaxOptTheta, vecLabel);
}
printf("%d classified images are saved into ResultImage/classNo directory\n",(int)vecImages.size());
for (i = 0;i<(int)vecImages.size();i++)
delete [] vecImages[i];
return;
}
