Skip to content

O
opencv
Commit f6daaccc authored by Suleyman TURKMEN's avatar Suleyman TURKMEN
Browse files
Options

Update train_HOG.cpp

parent 0e436c3f
Hide whitespace changes
Inline Side-by-side
Showing with 249 additions and 318 deletions
samples/cpp/train_HOG.cpp
View file @ f6daaccc
#include <opencv2/opencv.hpp> #include "opencv2/imgproc.hpp"
#include "opencv2/highgui.hpp"
#include "opencv2/ml.hpp"
#include "opencv2/objdetect.hpp"
#include <string>
#include <iostream> #include <iostream>
#include <fstream>
#include <vector>
#include <time.h> #include <time.h>
using namespace cv; using namespace cv;
using namespace cv::ml; using namespace cv::ml;
using namespace std; using namespace std;
void get_svm_detector(const Ptr<SVM>& svm, vector< float > & hog_detector ); void get_svm_detector( const Ptr< SVM > & svm, vector< float > & hog_detector );
void convert_to_ml(const std::vector< cv::Mat > & train_samples, cv::Mat& trainData ); void convert_to_ml( const std::vector< Mat > & train_samples, Mat& trainData );
void load_images( const string & prefix, const string & filename, vector< Mat > & img_lst ); void load_images( const String & dirname, vector< Mat > & img_lst, bool showImages );
void sample_neg( const vector< Mat > & full_neg_lst, vector< Mat > & neg_lst, const Size & size ); void sample_neg( const vector< Mat > & full_neg_lst, vector< Mat > & neg_lst, const Size & size );
Mat get_hogdescriptor_visu(const Mat& color_origImg, vector<float>& descriptorValues, const Size & size ); void computeHOGs( const Size wsize, const vector< Mat > & img_lst, vector< Mat > & gradient_lst );
void compute_hog( const vector< Mat > & img_lst, vector< Mat > & gradient_lst, const Size & size ); int test_trained_detector( String obj_det_filename, String test_dir, String videofilename );
void train_svm( const vector< Mat > & gradient_lst, const vector< int > & labels );
void draw_locations( Mat & img, const vector< Rect > & locations, const Scalar & color );
void test_it( const Size & size );
void get_svm_detector(const Ptr<SVM>& svm, vector< float > & hog_detector ) void get_svm_detector( const Ptr< SVM >& svm, vector< float > & hog_detector )
{ {
// get the support vectors // get the support vectors
Mat sv = svm->getSupportVectors(); Mat sv = svm->getSupportVectors();
const int sv_total = sv.rows; const int sv_total = sv.rows;
// get the decision function // get the decision function
Mat alpha, svidx; Mat alpha, svidx;
double rho = svm->getDecisionFunction(0, alpha, svidx); double rho = svm->getDecisionFunction( 0, alpha, svidx );
CV_Assert( alpha.total() == 1 && svidx.total() == 1 && sv_total == 1 ); CV_Assert( alpha.total() == 1 && svidx.total() == 1 && sv_total == 1 );
CV_Assert( (alpha.type() == CV_64F && alpha.at<double>(0) == 1.) || CV_Assert( (alpha.type() == CV_64F && alpha.at<double>(0) == 1.) ||
...@@ -37,70 +33,59 @@ void get_svm_detector(const Ptr<SVM>& svm, vector< float > & hog_detector ) ...@@ -37,70 +33,59 @@ void get_svm_detector(const Ptr<SVM>& svm, vector< float > & hog_detector )
hog_detector.clear(); hog_detector.clear();
hog_detector.resize(sv.cols + 1); hog_detector.resize(sv.cols + 1);
memcpy(&hog_detector[0], sv.ptr(), sv.cols*sizeof(hog_detector[0])); memcpy( &hog_detector[0], sv.ptr(), sv.cols*sizeof( hog_detector[0] ) );
hog_detector[sv.cols] = (float)-rho; hog_detector[sv.cols] = (float)-rho;
} }
/* /*
* Convert training/testing set to be used by OpenCV Machine Learning algorithms. * Convert training/testing set to be used by OpenCV Machine Learning algorithms.
* TrainData is a matrix of size (#samples x max(#cols,#rows) per samples), in 32FC1. * TrainData is a matrix of size (#samples x max(#cols,#rows) per samples), in 32FC1.
* Transposition of samples are made if needed. * Transposition of samples are made if needed.
*/ */
void convert_to_ml(const std::vector< cv::Mat > & train_samples, cv::Mat& trainData ) void convert_to_ml( const vector< Mat > & train_samples, Mat& trainData )
{ {
//--Convert data //--Convert data
const int rows = (int)train_samples.size(); const int rows = (int)train_samples.size();
const int cols = (int)std::max( train_samples[0].cols, train_samples[0].rows ); const int cols = (int)std::max( train_samples[0].cols, train_samples[0].rows );
cv::Mat tmp(1, cols, CV_32FC1); //< used for transposition if needed Mat tmp( 1, cols, CV_32FC1 ); //< used for transposition if needed
trainData = cv::Mat(rows, cols, CV_32FC1 ); trainData = Mat( rows, cols, CV_32FC1 );
vector< Mat >::const_iterator itr = train_samples.begin();
vector< Mat >::const_iterator end = train_samples.end(); for( size_t i = 0 ; i < train_samples.size(); ++i )
for( int i = 0 ; itr != end ; ++itr, ++i )
{ {
CV_Assert( itr->cols == 1 || CV_Assert( train_samples[i].cols == 1 || train_samples[i].rows == 1 );
itr->rows == 1 );
if( itr->cols == 1 ) if( train_samples[i].cols == 1 )
{ {
transpose( *(itr), tmp ); transpose( train_samples[i], tmp );
tmp.copyTo( trainData.row( i ) ); tmp.copyTo( trainData.row( (int)i ) );
} }
else if( itr->rows == 1 ) else if( train_samples[i].rows == 1 )
{ {
itr->copyTo( trainData.row( i ) ); train_samples[i].copyTo( trainData.row( (int)i ) );
} }
} }
} }
void load_images( const string & prefix, const string & filename, vector< Mat > & img_lst ) void load_images( const String & dirname, vector< Mat > & img_lst, bool showImages = false )
{ {
string line; vector< String > files;
ifstream file; glob( dirname, files );
file.open( (prefix+filename).c_str() ); for ( size_t i = 0; i < files.size(); ++i )
if( !file.is_open() )
{ {
cerr << "Unable to open the list of images from " << filename << " filename." << endl; Mat img = imread( files[i] ); // load the image
exit( -1 ); if ( img.empty() ) // invalid image, skip it.
} {
cout << files[i] << " is invalid!" << endl;
continue;
}
bool end_of_parsing = false; if ( showImages )
while( !end_of_parsing )
{
getline( file, line );
if( line.empty() ) // no more file to read
{ {
end_of_parsing = true; imshow( "image", img );
break; waitKey( 1 );
} }
Mat img = imread( (prefix+line).c_str() ); // load the image img_lst.push_back( img );
if( img.empty() ) // invalid image, just skip it.
continue;
#ifdef _DEBUG
imshow( "image", img );
waitKey( 10 );
#endif
img_lst.push_back( img.clone() );
} }
} }
...@@ -115,331 +100,277 @@ void sample_neg( const vector< Mat > & full_neg_lst, vector< Mat > & neg_lst, co ...@@ -115,331 +100,277 @@ void sample_neg( const vector< Mat > & full_neg_lst, vector< Mat > & neg_lst, co
srand( (unsigned int)time( NULL ) ); srand( (unsigned int)time( NULL ) );
vector< Mat >::const_iterator img = full_neg_lst.begin(); for ( size_t i = 0; i < full_neg_lst.size(); i++ )
vector< Mat >::const_iterator end = full_neg_lst.end();
for( ; img != end ; ++img )
{ {
box.x = rand() % (img->cols - size_x); box.x = rand() % ( full_neg_lst[i].cols - size_x );
box.y = rand() % (img->rows - size_y); box.y = rand() % ( full_neg_lst[i].rows - size_y );
Mat roi = (*img)(box); Mat roi = full_neg_lst[i]( box );
neg_lst.push_back( roi.clone() ); neg_lst.push_back( roi.clone() );
#ifdef _DEBUG
imshow( "img", roi.clone() );
waitKey( 10 );
#endif
} }
} }
// From http://www.juergenwiki.de/work/wiki/doku.php?id=public:hog_descriptor_computation_and_visualization void computeHOGs( const Size wsize, const vector< Mat > & img_lst, vector< Mat > & gradient_lst )
Mat get_hogdescriptor_visu(const Mat& color_origImg, vector<float>& descriptorValues, const Size & size )
{ {
const int DIMX = size.width; HOGDescriptor hog;
const int DIMY = size.height; hog.winSize = wsize;
float zoomFac = 3;
Mat visu;
resize(color_origImg, visu, Size( (int)(color_origImg.cols*zoomFac), (int)(color_origImg.rows*zoomFac) ) );
int cellSize = 8;
int gradientBinSize = 9;
float radRangeForOneBin = (float)(CV_PI/(float)gradientBinSize); // dividing 180 into 9 bins, how large (in rad) is one bin?
// prepare data structure: 9 orientation / gradient strenghts for each cell
int cells_in_x_dir = DIMX / cellSize;
int cells_in_y_dir = DIMY / cellSize;
float*** gradientStrengths = new float**[cells_in_y_dir];
int** cellUpdateCounter = new int*[cells_in_y_dir];
for (int y=0; y<cells_in_y_dir; y++)
{
gradientStrengths[y] = new float*[cells_in_x_dir];
cellUpdateCounter[y] = new int[cells_in_x_dir];
for (int x=0; x<cells_in_x_dir; x++)
{
gradientStrengths[y][x] = new float[gradientBinSize];
cellUpdateCounter[y][x] = 0;
for (int bin=0; bin<gradientBinSize; bin++)
gradientStrengths[y][x][bin] = 0.0;
}
}
// nr of blocks = nr of cells - 1 Rect r = Rect( 0, 0, wsize.width, wsize.height );
// since there is a new block on each cell (overlapping blocks!) but the last one r.x += ( img_lst[0].cols - r.width ) / 2;
int blocks_in_x_dir = cells_in_x_dir - 1; r.y += ( img_lst[0].rows - r.height ) / 2;
int blocks_in_y_dir = cells_in_y_dir - 1;
// compute gradient strengths per cell Mat gray;
int descriptorDataIdx = 0; vector< float > descriptors;
int cellx = 0;
int celly = 0;
for (int blockx=0; blockx<blocks_in_x_dir; blockx++) for( size_t i=0 ; i< img_lst.size(); i++ )
{ {
for (int blocky=0; blocky<blocks_in_y_dir; blocky++) cvtColor( img_lst[i](r), gray, COLOR_BGR2GRAY );
{ hog.compute( gray, descriptors, Size( 8, 8 ), Size( 0, 0 ) );
// 4 cells per block ... gradient_lst.push_back( Mat( descriptors ).clone() );
for (int cellNr=0; cellNr<4; cellNr++) }
{ }
// compute corresponding cell nr
cellx = blockx;
celly = blocky;
if (cellNr==1) celly++;
if (cellNr==2) cellx++;
if (cellNr==3)
{
cellx++;
celly++;
}
for (int bin=0; bin<gradientBinSize; bin++) int test_trained_detector( String obj_det_filename, String test_dir, String videofilename )
{ {
float gradientStrength = descriptorValues[ descriptorDataIdx ]; cout << "Testing trained detector..." << endl;
descriptorDataIdx++; HOGDescriptor hog;
hog.load( obj_det_filename );
gradientStrengths[celly][cellx][bin] += gradientStrength; vector< String > files;
glob( test_dir, files );
} // for (all bins) int delay = 0;
VideoCapture cap;
if ( videofilename != "" )
{
cap.open( videofilename );
}
// note: overlapping blocks lead to multiple updates of this sum! obj_det_filename = "testing " + obj_det_filename;
// we therefore keep track how often a cell was updated, namedWindow( obj_det_filename, WINDOW_NORMAL );
// to compute average gradient strengths
cellUpdateCounter[celly][cellx]++;
} // for (all cells) for( size_t i=0;; i++ )
{
Mat img;
if ( cap.isOpened() )
{
cap >> img;
delay = 1;
}
else if( i < files.size() )
{
img = imread( files[i] );
}
} // for (all block x pos) if ( img.empty() )
} // for (all block y pos) {
return 0;
}
vector< Rect > detections;
vector< double > foundWeights;
// compute average gradient strengths hog.detectMultiScale( img, detections, foundWeights );
for (celly=0; celly<cells_in_y_dir; celly++) for ( size_t j = 0; j < detections.size(); j++ )
{
for (cellx=0; cellx<cells_in_x_dir; cellx++)
{ {
Scalar color = Scalar( 0, foundWeights[j] * foundWeights[j] * 200, 0 );
rectangle( img, detections[j], color, img.cols / 400 + 1 );
}
float NrUpdatesForThisCell = (float)cellUpdateCounter[celly][cellx]; imshow( obj_det_filename, img );
// compute average gradient strenghts for each gradient bin direction if( 27 == waitKey( delay ) )
for (int bin=0; bin<gradientBinSize; bin++) {
{ return 0;
gradientStrengths[celly][cellx][bin] /= NrUpdatesForThisCell;
}
} }
} }
return 0;
}
// draw cells int main( int argc, char** argv )
for (celly=0; celly<cells_in_y_dir; celly++) {
const char* keys =
{ {
for (cellx=0; cellx<cells_in_x_dir; cellx++) "{help h| | show help message}"
{ "{pd | | path of directory contains possitive images}"
int drawX = cellx * cellSize; "{nd | | path of directory contains negative images}"
int drawY = celly * cellSize; "{td | | path of directory contains test images}"
"{tv | | test video file name}"
int mx = drawX + cellSize/2; "{dw | | width of the detector}"
int my = drawY + cellSize/2; "{dh | | height of the detector}"
"{d |false| train twice}"
rectangle(visu, Point((int)(drawX*zoomFac), (int)(drawY*zoomFac)), Point((int)((drawX+cellSize)*zoomFac), (int)((drawY+cellSize)*zoomFac)), Scalar(100,100,100), 1); "{t |false| test a trained detector}"
"{v |false| visualize training steps}"
// draw in each cell all 9 gradient strengths "{fn |my_detector.yml| file name of trained SVM}"
for (int bin=0; bin<gradientBinSize; bin++) };
{
float currentGradStrength = gradientStrengths[celly][cellx][bin]; CommandLineParser parser( argc, argv, keys );
// no line to draw? if ( parser.has( "help" ) )
if (currentGradStrength==0) {
continue; parser.printMessage();
exit( 0 );
float currRad = bin * radRangeForOneBin + radRangeForOneBin/2; }
float dirVecX = cos( currRad );
float dirVecY = sin( currRad );
float maxVecLen = (float)(cellSize/2.f);
float scale = 2.5; // just a visualization scale, to see the lines better
// compute line coordinates String pos_dir = parser.get< String >( "pd" );
float x1 = mx - dirVecX * currentGradStrength * maxVecLen * scale; String neg_dir = parser.get< String >( "nd" );
float y1 = my - dirVecY * currentGradStrength * maxVecLen * scale; String test_dir = parser.get< String >( "td" );
float x2 = mx + dirVecX * currentGradStrength * maxVecLen * scale; String obj_det_filename = parser.get< String >( "fn" );
float y2 = my + dirVecY * currentGradStrength * maxVecLen * scale; String videofilename = parser.get< String >( "tv" );
int detector_width = parser.get< int >( "dw" );
int detector_height = parser.get< int >( "dh" );
bool test_detector = parser.get< bool >( "t" );
bool train_twice = parser.get< bool >( "d" );
bool visualization = parser.get< bool >( "v" );
if ( test_detector )
{
test_trained_detector( obj_det_filename, test_dir, videofilename );
exit( 0 );
}
// draw gradient visualization if( pos_dir.empty() || neg_dir.empty() )
line(visu, Point((int)(x1*zoomFac),(int)(y1*zoomFac)), Point((int)(x2*zoomFac),(int)(y2*zoomFac)), Scalar(0,255,0), 1); {
parser.printMessage();
cout << "Wrong number of parameters.\n\n"
<< "Example command line:\n" << argv[0] << " -pd=/INRIAPerson/96X160H96/Train/pos -nd=/INRIAPerson/neg -td=/INRIAPerson/Test/pos -fn=HOGpedestrian96x160.yml -d\n"
<< "\nExample command line for testing trained detector:\n" << argv[0] << " -t -dw=96 -dh=160 -fn=HOGpedestrian96x160.yml -td=/INRIAPerson/Test/pos";
exit( 1 );
}
} // for (all bins) vector< Mat > pos_lst, full_neg_lst, neg_lst, gradient_lst;
vector< int > labels;
} // for (cellx) clog << "Positive images are being loaded..." ;
} // for (celly) load_images( pos_dir, pos_lst, visualization );
if ( pos_lst.size() > 0 )
{
clog << "...[done]" << endl;
}
else
{
clog << "no image in " << pos_dir <<endl;
return 1;
}
Size pos_image_size = pos_lst[0].size();
// don't forget to free memory allocated by helper data structures! for ( size_t i = 0; i < pos_lst.size(); ++i )
for (int y=0; y<cells_in_y_dir; y++)
{ {
for (int x=0; x<cells_in_x_dir; x++) if( pos_lst[i].size() != pos_image_size )
{ {
delete[] gradientStrengths[y][x]; cout << "All positive images should be same size!" << endl;
exit( 1 );
} }
delete[] gradientStrengths[y];
delete[] cellUpdateCounter[y];
} }
delete[] gradientStrengths;
delete[] cellUpdateCounter;
return visu; pos_image_size = pos_image_size / 8 * 8;
} // get_hogdescriptor_visu
void compute_hog( const vector< Mat > & img_lst, vector< Mat > & gradient_lst, const Size & size )
{
HOGDescriptor hog;
hog.winSize = size;
Mat gray;
vector< Point > location;
vector< float > descriptors;
vector< Mat >::const_iterator img = img_lst.begin(); if ( detector_width && detector_height )
vector< Mat >::const_iterator end = img_lst.end();
for( ; img != end ; ++img )
{ {
cvtColor( *img, gray, COLOR_BGR2GRAY ); pos_image_size = Size( detector_width, detector_height );
hog.compute( gray, descriptors, Size( 8, 8 ), Size( 0, 0 ), location );
gradient_lst.push_back( Mat( descriptors ).clone() );
#ifdef _DEBUG
imshow( "gradient", get_hogdescriptor_visu( img->clone(), descriptors, size ) );
waitKey( 10 );
#endif
} }
}
void train_svm( const vector< Mat > & gradient_lst, const vector< int > & labels ) labels.assign( pos_lst.size(), +1 );
{ const unsigned int old = (unsigned int)labels.size();
clog << "Negative images are being loaded...";
load_images( neg_dir, full_neg_lst, false );
sample_neg( full_neg_lst, neg_lst, pos_image_size );
clog << "...[done]" << endl;
labels.insert( labels.end(), neg_lst.size(), -1 );
CV_Assert( old < labels.size() );
clog << "Histogram of Gradients are being calculated for positive images...";
computeHOGs( pos_image_size, pos_lst, gradient_lst );
clog << "...[done]" << endl;
clog << "Histogram of Gradients are being calculated for negative images...";
computeHOGs( pos_image_size, neg_lst, gradient_lst );
clog << "...[done]" << endl;
Mat train_data; Mat train_data;
convert_to_ml( gradient_lst, train_data ); convert_to_ml( gradient_lst, train_data );
clog << "Start training..."; clog << "Training SVM...";
Ptr<SVM> svm = SVM::create(); Ptr< SVM > svm = SVM::create();
/* Default values to train SVM */ /* Default values to train SVM */
svm->setCoef0(0.0); svm->setCoef0( 0.0 );
svm->setDegree(3); svm->setDegree( 3 );
svm->setTermCriteria(TermCriteria( CV_TERMCRIT_ITER+CV_TERMCRIT_EPS, 1000, 1e-3 )); svm->setTermCriteria( TermCriteria( CV_TERMCRIT_ITER + CV_TERMCRIT_EPS, 1000, 1e-3 ) );
svm->setGamma(0); svm->setGamma( 0 );
svm->setKernel(SVM::LINEAR); svm->setKernel( SVM::LINEAR );
svm->setNu(0.5); svm->setNu( 0.5 );
svm->setP(0.1); // for EPSILON_SVR, epsilon in loss function? svm->setP( 0.1 ); // for EPSILON_SVR, epsilon in loss function?
svm->setC(0.01); // From paper, soft classifier svm->setC( 0.01 ); // From paper, soft classifier
svm->setType(SVM::EPS_SVR); // C_SVC; // EPSILON_SVR; // may be also NU_SVR; // do regression task svm->setType( SVM::EPS_SVR ); // C_SVC; // EPSILON_SVR; // may be also NU_SVR; // do regression task
svm->train(train_data, ROW_SAMPLE, Mat(labels)); svm->train( train_data, ROW_SAMPLE, Mat( labels ) );
clog << "...[done]" << endl; clog << "...[done]" << endl;
svm->save( "my_people_detector.yml" ); if ( train_twice )
}
void draw_locations( Mat & img, const vector< Rect > & locations, const Scalar & color )
{
if( !locations.empty() )
{ {
vector< Rect >::const_iterator loc = locations.begin(); clog << "Testing trained detector on negative images. This may take a few minutes...";
vector< Rect >::const_iterator end = locations.end(); HOGDescriptor my_hog;
for( ; loc != end ; ++loc ) my_hog.winSize = pos_image_size;
{
rectangle( img, *loc, color, 2 );
}
}
}
void test_it( const Size & size ) // Set the trained svm to my_hog
{ vector< float > hog_detector;
char key = 27; get_svm_detector( svm, hog_detector );
Scalar reference( 0, 255, 0 ); my_hog.setSVMDetector( hog_detector );
Scalar trained( 0, 0, 255 );
Mat img, draw;
Ptr<SVM> svm;
HOGDescriptor hog;
HOGDescriptor my_hog;
my_hog.winSize = size;
VideoCapture video;
vector< Rect > locations;
// Load the trained SVM.
svm = StatModel::load<SVM>( "my_people_detector.yml" );
// Set the trained svm to my_hog
vector< float > hog_detector;
get_svm_detector( svm, hog_detector );
my_hog.setSVMDetector( hog_detector );
// Set the people detector.
hog.setSVMDetector( hog.getDefaultPeopleDetector() );
// Open the camera.
video.open(0);
if( !video.isOpened() )
{
cerr << "Unable to open the device 0" << endl;
exit( -1 );
}
bool end_of_process = false; vector< Rect > detections;
while( !end_of_process ) vector< double > foundWeights;
{
video >> img;
if( img.empty() )
break;
draw = img.clone(); for ( size_t i = 0; i < full_neg_lst.size(); i++ )
{
my_hog.detectMultiScale( full_neg_lst[i], detections, foundWeights );
for ( size_t j = 0; j < detections.size(); j++ )
{
Mat detection = full_neg_lst[i]( detections[j] ).clone();
resize( detection, detection, pos_image_size );
neg_lst.push_back( detection );
}
if ( visualization )
{
for ( size_t j = 0; j < detections.size(); j++ )
{
rectangle( full_neg_lst[i], detections[j], Scalar( 0, 255, 0 ), 2 );
}
imshow( "testing trained detector on negative images", full_neg_lst[i] );
waitKey( 5 );
}
}
clog << "...[done]" << endl;
locations.clear(); labels.clear();
hog.detectMultiScale( img, locations ); labels.assign( pos_lst.size(), +1 );
draw_locations( draw, locations, reference ); labels.insert( labels.end(), neg_lst.size(), -1);
locations.clear(); gradient_lst.clear();
my_hog.detectMultiScale( img, locations ); clog << "Histogram of Gradients are being calculated for positive images...";
draw_locations( draw, locations, trained ); computeHOGs( pos_image_size, pos_lst, gradient_lst );
clog << "...[done]" << endl;
imshow( "Video", draw ); clog << "Histogram of Gradients are being calculated for negative images...";
key = (char)waitKey( 10 ); computeHOGs( pos_image_size, neg_lst, gradient_lst );
if( 27 == key ) clog << "...[done]" << endl;
end_of_process = true;
}
}
int main( int argc, char** argv ) clog << "Training SVM again...";
{ convert_to_ml( gradient_lst, train_data );
cv::CommandLineParser parser(argc, argv, "{help h|| show help message}" svm->train( train_data, ROW_SAMPLE, Mat( labels ) );
"{pd||pos_dir}{p||pos.lst}{nd||neg_dir}{n||neg.lst}"); clog << "...[done]" << endl;
if (parser.has("help"))
{
parser.printMessage();
exit(0);
} }
vector< Mat > pos_lst;
vector< Mat > full_neg_lst;
vector< Mat > neg_lst;
vector< Mat > gradient_lst;
vector< int > labels;
string pos_dir = parser.get<string>("pd");
string pos = parser.get<string>("p");
string neg_dir = parser.get<string>("nd");
string neg = parser.get<string>("n");
if( pos_dir.empty() || pos.empty() || neg_dir.empty() || neg.empty() )
{
cout << "Wrong number of parameters." << endl
<< "Usage: " << argv[0] << " --pd=pos_dir -p=pos.lst --nd=neg_dir -n=neg.lst" << endl
<< "example: " << argv[0] << " --pd=/INRIA_dataset/ -p=Train/pos.lst --nd=/INRIA_dataset/ -n=Train/neg.lst" << endl;
exit( -1 );
}
load_images( pos_dir, pos, pos_lst );
labels.assign( pos_lst.size(), +1 );
const unsigned int old = (unsigned int)labels.size();
load_images( neg_dir, neg, full_neg_lst );
sample_neg( full_neg_lst, neg_lst, Size( 96,160 ) );
labels.insert( labels.end(), neg_lst.size(), -1 );
CV_Assert( old < labels.size() );
compute_hog( pos_lst, gradient_lst, Size( 96, 160 ) );
compute_hog( neg_lst, gradient_lst, Size( 96, 160 ) );
train_svm( gradient_lst, labels ); vector< float > hog_detector;
get_svm_detector( svm, hog_detector );
HOGDescriptor hog;
hog.winSize = pos_image_size;
hog.setSVMDetector( hog_detector );
hog.save( obj_det_filename );
test_it( Size( 96, 160 ) ); // change with your parameters test_trained_detector( obj_det_filename, test_dir, videofilename );
return 0; return 0;
} }
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment