bool CascadeTrainI::createClassifier( const string& tempDir,
const string& vecFname, const string& bgName,
int numPos, int numNeg,
const TrainerPropertiesI *trainProps)
{
CvCascadeClassifier classifier;
int precalcValBufSize = 256,
precalcIdxBufSize = 256;
bool baseFormatSave = false;
CvCascadeParams cascadeParams;
cascadeParams.winSize.width = trainProps->windowSize.width;
cascadeParams.winSize.height = trainProps->windowSize.height;
CvCascadeBoostParams stageParams;
Ptr<CvFeatureParams> featureParams[] =
{ Ptr<CvFeatureParams>(new CvHaarFeatureParams),
Ptr<CvFeatureParams>(new CvLBPFeatureParams),
Ptr<CvFeatureParams>(new CvHOGFeatureParams)
};
cascadeParams.featureType = trainProps->featureType;
stageParams.boost_type = trainProps->boost_type;
bool res = classifier.train( tempDir,
vecFname,
bgName,
numPos, numNeg,
precalcValBufSize, precalcIdxBufSize,
trainProps->numStages,
cascadeParams,
*featureParams[cascadeParams.featureType],
stageParams,
baseFormatSave );
return res;
}
int main( int argc, char* argv[] )
{
CvCascadeClassifier classifier;
string cascadeDirName, vecName, bgName;
int numPos = 2000;
int numNeg = 1000;
int numStages = 20;
int numThreads = getNumThreads();
int precalcValBufSize = 256,
precalcIdxBufSize = 256;
bool baseFormatSave = false;
CvCascadeParams cascadeParams;
CvCascadeBoostParams stageParams;
Ptr<CvFeatureParams> featureParams[] = { makePtr<CvHaarFeatureParams>(),
makePtr<CvLBPFeatureParams>(),
makePtr<CvHOGFeatureParams>()
};
int fc = sizeof(featureParams)/sizeof(featureParams[0]);
if( argc == 1 )
{
cout << "Usage: " << argv[0] << endl;
cout << " -data <cascade_dir_name>" << endl;
cout << " -vec <vec_file_name>" << endl;
cout << " -bg <background_file_name>" << endl;
cout << " [-numPos <number_of_positive_samples = " << numPos << ">]" << endl;
cout << " [-numNeg <number_of_negative_samples = " << numNeg << ">]" << endl;
cout << " [-numStages <number_of_stages = " << numStages << ">]" << endl;
cout << " [-precalcValBufSize <precalculated_vals_buffer_size_in_Mb = " << precalcValBufSize << ">]" << endl;
cout << " [-precalcIdxBufSize <precalculated_idxs_buffer_size_in_Mb = " << precalcIdxBufSize << ">]" << endl;
cout << " [-baseFormatSave]" << endl;
cout << " [-numThreads <max_number_of_threads = " << numThreads << ">]" << endl;
cascadeParams.printDefaults();
stageParams.printDefaults();
for( int fi = 0; fi < fc; fi++ )
featureParams[fi]->printDefaults();
return 0;
}
for( int i = 1; i < argc; i++ )
{
bool set = false;
if( !strcmp( argv[i], "-data" ) )
{
cascadeDirName = argv[++i];
}
else if( !strcmp( argv[i], "-vec" ) )
{
vecName = argv[++i];
}
else if( !strcmp( argv[i], "-bg" ) )
{
bgName = argv[++i];
}
else if( !strcmp( argv[i], "-numPos" ) )
{
numPos = atoi( argv[++i] );
}
else if( !strcmp( argv[i], "-numNeg" ) )
{
numNeg = atoi( argv[++i] );
}
else if( !strcmp( argv[i], "-numStages" ) )
{
numStages = atoi( argv[++i] );
}
else if( !strcmp( argv[i], "-precalcValBufSize" ) )
{
precalcValBufSize = atoi( argv[++i] );
}
else if( !strcmp( argv[i], "-precalcIdxBufSize" ) )
{
precalcIdxBufSize = atoi( argv[++i] );
}
else if( !strcmp( argv[i], "-baseFormatSave" ) )
{
baseFormatSave = true;
}
else if( !strcmp( argv[i], "-numThreads" ) )
{
numThreads = atoi(argv[++i]);
}
else if ( cascadeParams.scanAttr( argv[i], argv[i+1] ) ) { i++; }
else if ( stageParams.scanAttr( argv[i], argv[i+1] ) ) { i++; }
else if ( !set )
{
for( int fi = 0; fi < fc; fi++ )
{
set = featureParams[fi]->scanAttr(argv[i], argv[i+1]);
if ( !set )
{
i++;
break;
}
}
}
}
setNumThreads( numThreads );
classifier.train( cascadeDirName,
vecName,
bgName,
numPos, numNeg,
precalcValBufSize, precalcIdxBufSize,
numStages,
cascadeParams,
*featureParams[cascadeParams.featureType],
stageParams,
baseFormatSave );
return 0;
}
int main( int argc, char* argv[] )
{
// переменные и классы
CvCascadeClassifier classifier; // from cascadeclassifier.h
string cascadeDirName, vecName, bgName;
int numPos = 2000; // Number of positive samples used in training for every classifier stage.
int numNeg = 1000; // Number of negative samples used in training for every classifier stage.
int numStages = 20; // Number of cascade stages to be trained.
int numThreads = getNumThreads(); // Maximum number of threads to use during training. Notice that the actual number of used
// threads may be lower, depending on your machine and compilation options.
// Function getNumThreads() from opencv2/core.hpp
int precalcValBufSize = 1024, // Size of buffer for precalculated feature values (in Mb).
precalcIdxBufSize = 1024; // Size of buffer for precalculated feature indices (in Mb). The more memory you have the
// faster the training process.
bool baseFormatSave = false; // This argument is actual in case of Haar-like features. If it is specified, the cascade will
// be saved in the old format.
double acceptanceRatioBreakValue = -1.0; // This argument is used to determine how precise your model should keep learning and
// when to stop. A good guideline is to train not further than 10e-5, to ensure
// the model does not overtrain on your training data. By default this value is
// set to -1 to disable this feature.
// Cascade parameters:
CvCascadeParams cascadeParams; // from cascadeclassifier.h
// здесь устанавливаются стандартные параметры каскада:
//
// CvCascadeParams::CvCascadeParams() : stageType( defaultStageType ), // Type of stages. Only boosted classifier are supported as a stage type at the moment.
// // -stageType <BOOST(default)>
// featureType( defaultFeatureType ), // Type of features: HAAR - Haar-like features,
// // LBP - local binary patterns.
// // -featureType<{HAAR(default), LBP}>
// winSize( cvSize(24, 24) ) // Size of training samples (in pixels). Must have exactly
// // the same values as used during training samples creation
// // (opencv_createsamples utility).
// {
// name = CC_CASCADE_PARAMS;
// }
// Boosted classifer parameters:
CvCascadeBoostParams stageParams; // from cascadeclassifier.h <-- boost.h
// CvCascadeBoostParams( int _boostType, // Type of boosted classifiers:
// // DAB - Discrete AdaBoost,
// // RAB - Real AdaBoost,
// // LB - LogitBoost,
// // GAB - Gentle AdaBoost.
// // -bt <{DAB, RAB, LB, GAB(default)}>
// float _minHitRate, // Minimal desired hit rate for each stage of the classifier. Overall hit rate may be estimated
// // as (min_hit_rate\^number_of_stages).
// // -minHitRate <min_hit_rate>
// float _maxFalseAlarm, // Maximal desired false alarm rate for each stage of the classifier. Overall false alarm rate
// // may be estimated as (max_false_alarm_rate\^number_of_stages).
// // -maxFalseAlarmRate <max_false_alarm_rate>
// double _weightTrimRate, // Specifies whether trimming should be used and its weight. A decent choice is 0.95.
// // -weightTrimRate <weight_trim_rate>
// int _maxDepth, // Maximal depth of a weak tree. A decent choice is 1, that is case of stumps.
// // -maxDepth <max_depth_of_weak_tree>
// int _maxWeakCount ); // Maximal count of weak trees for every cascade stage. The boosted classifier (stage) will
// // have so many weak trees (<=maxWeakCount), as needed to achieve the given -maxFalseAlarmRate.
// // -maxWeakCount <max_weak_tree_count>
// вот здесь непонятно, массив из фичей Haar, LBP и HOG!
Ptr<CvFeatureParams> featureParams[] = { makePtr<CvHaarFeatureParams>(),
makePtr<CvLBPFeatureParams>(),
makePtr<CvHOGFeatureParams>()
};
int fc = sizeof(featureParams)/sizeof(featureParams[0]); // вычисление размера массива
// здесь проиходит вывод справочной инфы, если не были переданы параметры
if( argc == 1 )
{
cout << "Usage: " << argv[0] << endl;
cout << " -data <cascade_dir_name>" << endl;
cout << " -vec <vec_file_name>" << endl;
cout << " -bg <background_file_name>" << endl;
cout << " [-numPos <number_of_positive_samples = " << numPos << ">]" << endl;
cout << " [-numNeg <number_of_negative_samples = " << numNeg << ">]" << endl;
cout << " [-numStages <number_of_stages = " << numStages << ">]" << endl;
cout << " [-precalcValBufSize <precalculated_vals_buffer_size_in_Mb = " << precalcValBufSize << ">]" << endl;
cout << " [-precalcIdxBufSize <precalculated_idxs_buffer_size_in_Mb = " << precalcIdxBufSize << ">]" << endl;
cout << " [-baseFormatSave]" << endl;
cout << " [-numThreads <max_number_of_threads = " << numThreads << ">]" << endl;
cout << " [-acceptanceRatioBreakValue <value> = " << acceptanceRatioBreakValue << ">]" << endl;
cascadeParams.printDefaults();
stageParams.printDefaults();
for( int fi = 0; fi < fc; fi++ )
featureParams[fi]->printDefaults();
return 0;
}
// распознавание аргументов командной строки Common arguments:
// -data <cascade_dir_name>
// -vec <vec_file_name>
// -bg <background_file_name>
// -numPos <number_of_positive_samples>
// -numNeg <number_of_negative_samples>
// -numStages <number_of_stages>
// -precalcValBufSize <precalculated_vals_buffer_size_in_Mb>
// -precalcIdxBufSize <precalculated_idxs_buffer_size_in_Mb>
// -baseFormatSave
// -numThreads <max_number_of_threads>
// -acceptanceRatioBreakValue <break_value>
for( int i = 1; i < argc; i++ )
{
bool set = false;
if( !strcmp( argv[i], "-data" ) )
{
cascadeDirName = argv[++i];
}
else if( !strcmp( argv[i], "-vec" ) )
{
vecName = argv[++i];
}
else if( !strcmp( argv[i], "-bg" ) )
{
bgName = argv[++i];
}
else if( !strcmp( argv[i], "-numPos" ) )
{
numPos = atoi( argv[++i] );
}
else if( !strcmp( argv[i], "-numNeg" ) )
{
numNeg = atoi( argv[++i] );
}
else if( !strcmp( argv[i], "-numStages" ) )
{
numStages = atoi( argv[++i] );
}
else if( !strcmp( argv[i], "-precalcValBufSize" ) )
{
precalcValBufSize = atoi( argv[++i] );
}
else if( !strcmp( argv[i], "-precalcIdxBufSize" ) )
{
precalcIdxBufSize = atoi( argv[++i] );
}
else if( !strcmp( argv[i], "-baseFormatSave" ) )
{
baseFormatSave = true;
}
else if( !strcmp( argv[i], "-numThreads" ) )
{
numThreads = atoi(argv[++i]);
}
else if( !strcmp( argv[i], "-acceptanceRatioBreakValue" ) )
{
acceptanceRatioBreakValue = atof(argv[++i]);
}
// распознавание аргументов командной строки Cascade parameters:
else if ( cascadeParams.scanAttr( argv[i], argv[i+1] ) ) { i++; }
// распознавание аргументов командной строки Boosted classifer parameters:
else if ( stageParams.scanAttr( argv[i], argv[i+1] ) ) { i++; }
// что-то происходит, если ещё какие параметры идут, возможно, Haar-like feature parameters:
// -mode <BASIC (default) | CORE | ALL>
// Selects the type of Haar features set used in training. BASIC use only upright features, while ALL uses
// the full set of upright and 45 degree rotated feature set. See [86] for more details.
else if ( !set )
{
for( int fi = 0; fi < fc; fi++ )
{
set = featureParams[fi]->scanAttr(argv[i], argv[i+1]);
if ( !set )
{
i++;
break;
}
}
}
}
// set max_number_of_threads
// Maximum number of threads to use during training. Notice that the actual number of used
// threads may be lower, depending on your machine and compilation options.
setNumThreads( numThreads );
// метод train определён в cascadeclassifier.h, он тренирует классификатор и сохраняет его в xml-file
// After the traincascade application has finished its work, the trained cascade will be saved
// in cascade.xml file in the folder, which was passed as -data parameter. Other files in this folder
// are created for the case of interrupted training, so you may delete them after completion of training.
classifier.train( cascadeDirName,
vecName,
bgName,
numPos, numNeg,
precalcValBufSize, precalcIdxBufSize,
numStages,
cascadeParams,
*featureParams[cascadeParams.featureType],
stageParams,
baseFormatSave,
acceptanceRatioBreakValue );
return 0;
}
