void Build_Vocabulary::doKmeansAll(String postFix){
clock_t start, finish;
double elapsed_time;
start = time(NULL);
cout << "Loading Descriptor Files..." << endl;
FileStorage fs("training_descriptors_bike_" +postFix+ ".txt", FileStorage::READ);
Mat training_descriptors_bike;
fs["training_descriptors"] >> training_descriptors_bike;
fs.release();
FileStorage fs1("training_descriptors_cars_" +postFix+ ".txt", FileStorage::READ);
Mat training_descriptors_cars;
fs1["training_descriptors"] >> training_descriptors_cars;
fs1.release();
FileStorage fs2("training_descriptors_person_" +postFix+ ".txt", FileStorage::READ);
Mat training_descriptors_person;
fs2["training_descriptors"] >> training_descriptors_person;
fs2.release();
FileStorage fs3("training_descriptors_none_" +postFix+ ".txt", FileStorage::READ);
Mat training_descriptors_none;
fs3["training_descriptors"] >> training_descriptors_none;
fs3.release();
/* */
BOWKMeansTrainer bowtrainer(1000); //num clusters
bowtrainer.add(training_descriptors_bike);
bowtrainer.add(training_descriptors_person);
bowtrainer.add(training_descriptors_cars);
bowtrainer.add(training_descriptors_none);
cout << "Clustering Bag-of-Words features..." << endl;
Mat vocabulary = bowtrainer.cluster();
FileStorage fs4("vocabulary_1000_" +postFix+ ".txt", FileStorage::WRITE);
fs4 << "vocabulary" << vocabulary;
fs4.release();
/* */
finish=time(NULL);
elapsed_time = finish-start;
int hours = (int) elapsed_time / 3600;
int minutes = (int) (elapsed_time - hours * 3600) / 60;
int seconds = (int) elapsed_time - hours * 3600 - minutes * 60;
cout << "Elapsed Time for Clustering: " << hours << ":" << minutes << ":" << seconds << endl << endl;
cout << "------- Finished Build Vocabulary ---------\n" << endl;
}
void Build_Vocabulary::doKmeans(String filename, String itemname){
clock_t start, finish;
double elapsed_time;
start = time(NULL);
cout << "Loading Descriptor Files..." << endl;
FileStorage fs("training_descriptors_" + filename + ".txt", FileStorage::READ);
Mat training_descriptors;
fs["training_descriptors"] >> training_descriptors;
fs.release();
/* */
BOWKMeansTrainer bowtrainer(1000); //num clusters
bowtrainer.add(training_descriptors);
cout << "Clustering Bag-of-Words features..." << endl;
Mat vocabulary = bowtrainer.cluster();
FileStorage fs1("vocabulary_" + filename + "_1000.txt", FileStorage::WRITE);
fs1 << "vocabulary" << vocabulary;
fs1.release();
/*
Mat Label, kMeanCenter;
TermCriteria KmeanTerm(CV_TERMCRIT_EPS, 100, 0.001);
kmeans(training_descriptors, 1000, Label , KmeanTerm, 2, KMEANS_PP_CENTERS, kMeanCenter);
*/
/*
FileStorage fs2("kMeans_1000.txt", FileStorage::WRITE);
fs2 << "label" << Label;
fs2 << "kMeanCenter" << kMeanCenter;
fs2.release();
*/
/* */
finish=time(NULL);
elapsed_time = finish-start;
int hours = (int) elapsed_time / 3600;
int minutes = (int) (elapsed_time - hours * 3600) / 60;
int seconds = (int) elapsed_time - hours * 3600 - minutes * 60;
cout << "Elapsed Time for Clustering: " << hours << ":" << minutes << ":" << seconds << endl << endl;
cout << "------- Finished Build Vocabulary ---------\n" << endl;
}
void VisualBOW::createVocabulary(Mat &descriptors)
{
BOWKMeansTrainer bowtrainer(1000);
vocabulary = bowtrainer.cluster(descriptors);
}
int main(int argc, char** argv) {
string dir = "foodcamimages/TRAIN", filepath;
DIR *dp;
struct dirent *dirp;
struct stat filestat;
dp = opendir( dir.c_str() );
// detecting keypoints
SurfFeatureDetector detector(400);
//FastFeatureDetector detector(1,true);
vector<KeyPoint> keypoints;
// computing descriptors
//Ptr<DescriptorExtractor > extractor(new SurfDescriptorExtractor());// extractor;
Ptr<DescriptorExtractor > extractor(
new OpponentColorDescriptorExtractor(
Ptr<DescriptorExtractor>(new SurfDescriptorExtractor())
)
);
Mat descriptors;
Mat training_descriptors(1,extractor->descriptorSize(),extractor->descriptorType());
Mat img;
cout << "------- build vocabulary ---------\n";
cout << "extract descriptors.."<<endl;
//int count = 0;
Rect clipping_rect = Rect(0,120,640,480-120);
Mat bg_ = imread("background.png")(clipping_rect), img_fg;
while (dirp = readdir( dp ))
{
// count++;
filepath = dir + "/" + dirp->d_name;
// If the file is a directory (or is in some way invalid) we'll skip it
if (stat( filepath.c_str(), &filestat )) continue;
if (S_ISDIR( filestat.st_mode )) continue;
img = imread(filepath);
if (!img.data) {
continue;
}
img = img(clipping_rect);
img_fg = img - bg_;
detector.detect(img_fg, keypoints);
// {
// Mat out; //img_fg.copyTo(out);
// drawKeypoints(img, keypoints, out, Scalar(255));
// imshow("fg",img_fg);
// imshow("keypoints", out);
// waitKey(0);
// }
extractor->compute(img, keypoints, descriptors);
training_descriptors.push_back(descriptors);
cout << ".";
}
cout << endl;
closedir( dp );
cout << "Total descriptors: " << training_descriptors.rows << endl;
FileStorage fs("training_descriptors.yml", FileStorage::WRITE);
fs << "training_descriptors" << training_descriptors;
fs.release();
BOWKMeansTrainer bowtrainer(1000); //num clusters
bowtrainer.add(training_descriptors);
cout << "cluster BOW features" << endl;
Mat vocabulary = bowtrainer.cluster();
FileStorage fs1("vocabulary_color_1000.yml", FileStorage::WRITE);
fs1 << "vocabulary" << vocabulary;
fs1.release();
}
