void GLinearRegressor_linear_test(GRand& prng)
{
// Train
GMatrix features1(0, 3);
GMatrix labels1(0, 1);
for(size_t i = 0; i < 1000; i++)
{
GVec& vec = features1.newRow();
vec[0] = prng.uniform();
vec[1] = prng.uniform(); // irrelevant attribute
vec[2] = prng.uniform();
labels1.newRow()[0] = 0.3 * vec[0] + 2.0 * vec[2] + 5.0;
}
GLinearRegressor lr;
lr.train(features1, labels1);
// Check some values
if(lr.beta()->rows() != 1 || lr.beta()->cols() != 3)
throw Ex("failed");
if(std::abs(lr.beta()->row(0)[0] - 0.3) > 1e-6)
throw Ex("failed");
if(std::abs(lr.beta()->row(0)[1] - 0.0) > 1e-6)
throw Ex("failed");
if(std::abs(lr.beta()->row(0)[2] - 2.0) > 1e-6)
throw Ex("failed");
if(std::abs(lr.epsilon()[0] - 5.0) > 1e-6)
throw Ex("failed");
// Test
GMatrix features2(0, 3);
GMatrix labels2(0, 1);
for(size_t i = 0; i < 1000; i++)
{
GVec& vec = features2.newRow();
vec[0] = prng.uniform();
vec[1] = prng.uniform(); // irrelevant attribute
vec[2] = prng.uniform();
labels2.newRow()[0] = 0.3 * vec[0] + 2.0 * vec[2] + 5.0;
}
double rmse = sqrt(lr.sumSquaredError(features2, labels2) / features2.rows());
if(rmse > 0.0224)
throw Ex("failed");
}
int main(int argc, char *argv[])
{
if (argc != 4)
{
cout << "Usage: " << endl;
cout << "ACCV2009.exe Folder imfn1 imfn2" << endl;
return -1;
}
string folder = argv[1];
string imfn1 = argv[2];
string imfn2 = argv[3];
string outfolder = "../output";
_mkdir(outfolder.c_str());
Mat im1 = imread(folder + imfn1 + ".png", 1);
Mat im2 = imread(folder + imfn2 + ".png", 1);
int width = im1.cols;
int height = im1.rows;
//颜色空间转换
Mat im1_H, im1_S, im1_I;
Mat im2_H, im2_S, im2_I;
//CvtColorBGR2HSI(im1, im1_H, im1_S, im1_I);
//CvtColorBGR2HSI(im2, im2_H, im2_S, im2_I);
cv_CvtColorBGR2HSI(im1, im1_H, im1_S, im1_I);
cv_CvtColorBGR2HSI(im2, im2_H, im2_S, im2_I);
int sigmaS = 5;
float sigmaR = 10;
int minR = 800;
vector<int> labels1(0), labels2(0);
//#define RQ_DEBUG
#ifdef RQ_DEBUG
int regionum1, regionum2;
DoMeanShift(im1, sigmaS, sigmaR, minR, labels1, regionum1);
DoMeanShift(im2, sigmaS, sigmaR, minR, labels2, regionum2);
cout << endl;
saveLabels(labels1, width, height, outfolder+"/labels_" + imfn1 + ".txt");
saveLabels(labels2, width, height, outfolder+"/labels_" + imfn2 + ".txt");
#else
int regionum1, regionum2;
readLabels(outfolder+"/labels_" + imfn1 + ".txt", width, height, labels1, regionum1);
readLabels(outfolder+"/labels_" + imfn2 + ".txt", width, height, labels2, regionum2);
#endif
string siftmatchfn = folder + "matches_sift.txt";
vector<Point2f> features1(0), features2(0);
ReadSiftMatches(siftmatchfn, features1, features2);
vector<vector<Point2f>> matches1(0), matches2(0);
vector<vector<Point2f>> pixelTable2(0);
ComputeRegionMatches(labels2, regionum2, width, features1, features2, matches1, matches2);
cout << "all regions have matches." << endl << endl;
//显示每个区域的matches : 显示一次就好
vector<float> DeltaH(0), DeltaS(0), DeltaI(0);
RegionDeltaColor(im1_H, im2_H, matches1, matches2, DeltaH);
RegionDeltaColor(im1_S, im2_S, matches1, matches2, DeltaS);
RegionDeltaColor(im1_I, im2_I, matches1, matches2, DeltaI);
Mat new_im2_H, new_im2_S, new_im2_I;
CorrectColor(im2_H, labels2, DeltaH, new_im2_H);
CorrectColor(im2_S, labels2, DeltaS, new_im2_S);
CorrectColor(im2_I, labels2, DeltaI, new_im2_I);
Mat new_im2;
//CvtColorHSI2BGR(new_im2_H, new_im2_S, new_im2_I, new_im2);
cv_CVtColorHSI2BGR(new_im2_H, new_im2_S, new_im2_I, new_im2);
cout << "done." << endl;
imshow("new im2", new_im2);
waitKey(0);
destroyAllWindows();
string savefn = outfolder + "/accv2009_" + imfn2 + ".png";
cout << "save " << savefn << endl;
imwrite(savefn, new_im2);
}
void classify(){
cv::HOGDescriptor hog, hog2;
cv::Mat image, im;
std::string name;
const cv::Size trainingPadding = cv::Size(0,0);
const cv::Size winStride = cv::Size(8,8);//Strid 16= no overlap
std::vector<cv::Point> locations;
std::vector<float> featureVector, featureVector2;
std::vector<int> voting(3);
CvSVM SVM_pen, SVM_scissors, SVM_pen_scissors;
float classification_result1, classification_result2, classification_result3;
cv::Mat circle= cv::Mat::zeros(cv::Size(RES_X,RES_Y),CV_8U);
int thickness=-1, lineType=8;
int max;
cv::circle(circle, cv::Size(circle.cols/2, circle.rows/2),circle.rows/2, 255, thickness, lineType );
//filling vector for voting
std::fill( voting.begin(), voting.end(), 0 );
hog.blockSize= cv::Size(16,16);
hog.winSize= cv::Size(RES_X,RES_Y);
hog.blockStride=cv::Size(8,8);
hog.cellSize= cv::Size(8,8);
hog2.blockSize= cv::Size(16,16);
hog2.winSize= cv::Size(RES_X,RES_Y);
hog2.blockStride=cv::Size(8,8);
hog2.cellSize= cv::Size(16,16);
//Loading support vector machines
SVM_scissors.load("svm_scissors");
SVM_pen.load("svm_pen");
SVM_pen_scissors.load("pen_scissors_svm");
image=extracted_image;
cv::resize(image, im, cv::Size(RES_X,RES_Y));
im.copyTo(image, circle);
//Finding features
hog.compute(image, featureVector, winStride, trainingPadding, locations);
hog2.compute(image, featureVector2, winStride, trainingPadding, locations);
cv::Mat features(1, featureVector.size(), CV_32F);
cv::Mat features2(1, featureVector2.size(), CV_32F);
for(int j=0; j<featureVector.size(); j++){
features.at<float>(0,j)= featureVector.at(j);
}
for(int j=0; j<featureVector2.size(); j++){
features2.at<float>(0,j)= featureVector2.at(j);
}
//Evaluating features in the support vector machine
classification_result1= SVM_scissors.predict(features);
classification_result2= SVM_pen.predict(features);
classification_result3= SVM_pen_scissors.predict(features2);
std::cout<<"Output of scissors classifier:"<<classification_result1<<std::endl;
std::cout<<"Output of pen classifier:"<<classification_result2<<std::endl;
std::cout<<"Output of scissors/pen classifier:"<<classification_result3<<std::endl;
//Deciding the classification result
classification_result3= SVM_pen_scissors.predict(features2);
if( classification_result3==1){
classification_result1= SVM_scissors.predict(features);
if(classification_result1==1) {
std::cout<<"IT IS A SCISSORS."<<std::endl;
c= SCISSORS;
}
else{
std::cout<<"IT IS NEITHER A SCISSOPRS NOR A PEN"<<std::endl;
c= NO_IDENTIFIED;
}
}
else {
classification_result2= SVM_pen.predict(features);
if(classification_result2==1) {
std::cout<<"IT IS A PEN."<<std::endl;
c= PEN;
}
else{
std::cout<<"IT IS NEITHER A SCISSORS NOR A PEN."<<std::endl;
c= NO_IDENTIFIED;
}
}
}