potok/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2026-07-30 07:43:03 +04:00
Code Packages Releases Wiki Activity

initial commit; ml has been refactored; it compiles and the tests run well; some other modules, apps and samples do not compile; to be fixed

Browse Source
This commit is contained in:
Vadim Pisarevsky
2014-07-29 23:54:23 +04:00
parent dce1824a91
commit ba3783d205
25 changed files with 8320 additions and 21792 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/ml/test/test_mltests2.cpp
+161 -476
View File
@@ -44,257 +44,49 @@
using namespace cv;
using namespace std;
// auxiliary functions
// 1. nbayes
void nbayes_check_data( CvMLData* _data )
{
if( _data->get_missing() )
CV_Error( CV_StsBadArg, "missing values are not supported" );
const CvMat* var_types = _data->get_var_types();
bool is_classifier = var_types->data.ptr[var_types->cols-1] == CV_VAR_CATEGORICAL;
Mat _var_types = cvarrToMat(var_types);
if( ( fabs( cvtest::norm( _var_types, Mat::zeros(_var_types.dims, _var_types.size, _var_types.type()), CV_L1 ) -
(var_types->rows + var_types->cols - 2)*CV_VAR_ORDERED - CV_VAR_CATEGORICAL ) > FLT_EPSILON ) ||
!is_classifier )
CV_Error( CV_StsBadArg, "incorrect types of predictors or responses" );
}
bool nbayes_train( CvNormalBayesClassifier* nbayes, CvMLData* _data )
{
nbayes_check_data( _data );
const CvMat* values = _data->get_values();
const CvMat* responses = _data->get_responses();
const CvMat* train_sidx = _data->get_train_sample_idx();
const CvMat* var_idx = _data->get_var_idx();
return nbayes->train( values, responses, var_idx, train_sidx );
}
float nbayes_calc_error( CvNormalBayesClassifier* nbayes, CvMLData* _data, int type, vector<float> *resp )
{
float err = 0;
nbayes_check_data( _data );
const CvMat* values = _data->get_values();
const CvMat* response = _data->get_responses();
const CvMat* sample_idx = (type == CV_TEST_ERROR) ? _data->get_test_sample_idx() : _data->get_train_sample_idx();
int* sidx = sample_idx ? sample_idx->data.i : 0;
int r_step = CV_IS_MAT_CONT(response->type) ?
1 : response->step / CV_ELEM_SIZE(response->type);
int sample_count = sample_idx ? sample_idx->cols : 0;
sample_count = (type == CV_TRAIN_ERROR && sample_count == 0) ? values->rows : sample_count;
float* pred_resp = 0;
if( resp && (sample_count > 0) )
{
resp->resize( sample_count );
pred_resp = &((*resp)[0]);
}
for( int i = 0; i < sample_count; i++ )
{
CvMat sample;
int si = sidx ? sidx[i] : i;
cvGetRow( values, &sample, si );
float r = (float)nbayes->predict( &sample, 0 );
if( pred_resp )
pred_resp[i] = r;
int d = fabs((double)r - response->data.fl[si*r_step]) <= FLT_EPSILON ? 0 : 1;
err += d;
}
err = sample_count ? err / (float)sample_count * 100 : -FLT_MAX;
return err;
}
// 2. knearest
void knearest_check_data_and_get_predictors( CvMLData* _data, CvMat* _predictors )
{
const CvMat* values = _data->get_values();
const CvMat* var_idx = _data->get_var_idx();
if( var_idx->cols + var_idx->rows != values->cols )
CV_Error( CV_StsBadArg, "var_idx is not supported" );
if( _data->get_missing() )
CV_Error( CV_StsBadArg, "missing values are not supported" );
int resp_idx = _data->get_response_idx();
if( resp_idx == 0)
cvGetCols( values, _predictors, 1, values->cols );
else if( resp_idx == values->cols - 1 )
cvGetCols( values, _predictors, 0, values->cols - 1 );
else
CV_Error( CV_StsBadArg, "responses must be in the first or last column; other cases are not supported" );
}
bool knearest_train( CvKNearest* knearest, CvMLData* _data )
{
const CvMat* responses = _data->get_responses();
const CvMat* train_sidx = _data->get_train_sample_idx();
bool is_regression = _data->get_var_type( _data->get_response_idx() ) == CV_VAR_ORDERED;
CvMat predictors;
knearest_check_data_and_get_predictors( _data, &predictors );
return knearest->train( &predictors, responses, train_sidx, is_regression );
}
float knearest_calc_error( CvKNearest* knearest, CvMLData* _data, int k, int type, vector<float> *resp )
{
float err = 0;
const CvMat* response = _data->get_responses();
const CvMat* sample_idx = (type == CV_TEST_ERROR) ? _data->get_test_sample_idx() : _data->get_train_sample_idx();
int* sidx = sample_idx ? sample_idx->data.i : 0;
int r_step = CV_IS_MAT_CONT(response->type) ?
1 : response->step / CV_ELEM_SIZE(response->type);
bool is_regression = _data->get_var_type( _data->get_response_idx() ) == CV_VAR_ORDERED;
CvMat predictors;
knearest_check_data_and_get_predictors( _data, &predictors );
int sample_count = sample_idx ? sample_idx->cols : 0;
sample_count = (type == CV_TRAIN_ERROR && sample_count == 0) ? predictors.rows : sample_count;
float* pred_resp = 0;
if( resp && (sample_count > 0) )
{
resp->resize( sample_count );
pred_resp = &((*resp)[0]);
}
if ( !is_regression )
{
for( int i = 0; i < sample_count; i++ )
{
CvMat sample;
int si = sidx ? sidx[i] : i;
cvGetRow( &predictors, &sample, si );
float r = knearest->find_nearest( &sample, k );
if( pred_resp )
pred_resp[i] = r;
int d = fabs((double)r - response->data.fl[si*r_step]) <= FLT_EPSILON ? 0 : 1;
err += d;
}
err = sample_count ? err / (float)sample_count * 100 : -FLT_MAX;
}
else
{
for( int i = 0; i < sample_count; i++ )
{
CvMat sample;
int si = sidx ? sidx[i] : i;
cvGetRow( &predictors, &sample, si );
float r = knearest->find_nearest( &sample, k );
if( pred_resp )
pred_resp[i] = r;
float d = r - response->data.fl[si*r_step];
err += d*d;
}
err = sample_count ? err / (float)sample_count : -FLT_MAX;
}
return err;
}
// 3. svm
int str_to_svm_type(String& str)
{
if( !str.compare("C_SVC") )
return CvSVM::C_SVC;
return SVM::C_SVC;
if( !str.compare("NU_SVC") )
return CvSVM::NU_SVC;
return SVM::NU_SVC;
if( !str.compare("ONE_CLASS") )
return CvSVM::ONE_CLASS;
return SVM::ONE_CLASS;
if( !str.compare("EPS_SVR") )
return CvSVM::EPS_SVR;
return SVM::EPS_SVR;
if( !str.compare("NU_SVR") )
return CvSVM::NU_SVR;
return SVM::NU_SVR;
CV_Error( CV_StsBadArg, "incorrect svm type string" );
return -1;
}
int str_to_svm_kernel_type( String& str )
{
if( !str.compare("LINEAR") )
return CvSVM::LINEAR;
return SVM::LINEAR;
if( !str.compare("POLY") )
return CvSVM::POLY;
return SVM::POLY;
if( !str.compare("RBF") )
return CvSVM::RBF;
return SVM::RBF;
if( !str.compare("SIGMOID") )
return CvSVM::SIGMOID;
return SVM::SIGMOID;
CV_Error( CV_StsBadArg, "incorrect svm type string" );
return -1;
}
void svm_check_data( CvMLData* _data )
Ptr<SVM> svm_train_auto( Ptr<TrainData> _data, SVM::Params _params,
int k_fold, ParamGrid C_grid, ParamGrid gamma_grid,
ParamGrid p_grid, ParamGrid nu_grid, ParamGrid coef_grid,
ParamGrid degree_grid )
{
if( _data->get_missing() )
CV_Error( CV_StsBadArg, "missing values are not supported" );
const CvMat* var_types = _data->get_var_types();
for( int i = 0; i < var_types->cols-1; i++ )
if (var_types->data.ptr[i] == CV_VAR_CATEGORICAL)
{
char msg[50];
sprintf( msg, "incorrect type of %d-predictor", i );
CV_Error( CV_StsBadArg, msg );
}
}
bool svm_train( CvSVM* svm, CvMLData* _data, CvSVMParams _params )
{
svm_check_data(_data);
const CvMat* _train_data = _data->get_values();
const CvMat* _responses = _data->get_responses();
const CvMat* _var_idx = _data->get_var_idx();
const CvMat* _sample_idx = _data->get_train_sample_idx();
return svm->train( _train_data, _responses, _var_idx, _sample_idx, _params );
}
bool svm_train_auto( CvSVM* svm, CvMLData* _data, CvSVMParams _params,
int k_fold, CvParamGrid C_grid, CvParamGrid gamma_grid,
CvParamGrid p_grid, CvParamGrid nu_grid, CvParamGrid coef_grid,
CvParamGrid degree_grid )
{
svm_check_data(_data);
const CvMat* _train_data = _data->get_values();
const CvMat* _responses = _data->get_responses();
const CvMat* _var_idx = _data->get_var_idx();
const CvMat* _sample_idx = _data->get_train_sample_idx();
return svm->train_auto( _train_data, _responses, _var_idx,
_sample_idx, _params, k_fold, C_grid, gamma_grid, p_grid, nu_grid, coef_grid, degree_grid );
}
float svm_calc_error( CvSVM* svm, CvMLData* _data, int type, vector<float> *resp )
{
svm_check_data(_data);
float err = 0;
const CvMat* values = _data->get_values();
const CvMat* response = _data->get_responses();
const CvMat* sample_idx = (type == CV_TEST_ERROR) ? _data->get_test_sample_idx() : _data->get_train_sample_idx();
const CvMat* var_types = _data->get_var_types();
int* sidx = sample_idx ? sample_idx->data.i : 0;
int r_step = CV_IS_MAT_CONT(response->type) ?
1 : response->step / CV_ELEM_SIZE(response->type);
bool is_classifier = var_types->data.ptr[var_types->cols-1] == CV_VAR_CATEGORICAL;
int sample_count = sample_idx ? sample_idx->cols : 0;
sample_count = (type == CV_TRAIN_ERROR && sample_count == 0) ? values->rows : sample_count;
float* pred_resp = 0;
if( resp && (sample_count > 0) )
{
resp->resize( sample_count );
pred_resp = &((*resp)[0]);
}
if ( is_classifier )
{
for( int i = 0; i < sample_count; i++ )
{
CvMat sample;
int si = sidx ? sidx[i] : i;
cvGetRow( values, &sample, si );
float r = svm->predict( &sample );
if( pred_resp )
pred_resp[i] = r;
int d = fabs((double)r - response->data.fl[si*r_step]) <= FLT_EPSILON ? 0 : 1;
err += d;
}
err = sample_count ? err / (float)sample_count * 100 : -FLT_MAX;
}
else
{
for( int i = 0; i < sample_count; i++ )
{
CvMat sample;
int si = sidx ? sidx[i] : i;
cvGetRow( values, &sample, si );
float r = svm->predict( &sample );
if( pred_resp )
pred_resp[i] = r;
float d = r - response->data.fl[si*r_step];
err += d*d;
}
err = sample_count ? err / (float)sample_count : -FLT_MAX;
}
return err;
Mat _train_data = _data->getSamples();
Mat _responses = _data->getResponses();
Mat _var_idx = _data->getVarIdx();
Mat _sample_idx = _data->getTrainSampleIdx();
Ptr<SVM> svm = SVM::create(_params);
if( svm->trainAuto( _data, k_fold, C_grid, gamma_grid, p_grid, nu_grid, coef_grid, degree_grid ) )
return svm;
return Ptr<SVM>();
}
// 4. em
@@ -302,79 +94,66 @@ float svm_calc_error( CvSVM* svm, CvMLData* _data, int type, vector<float> *resp
int str_to_ann_train_method( String& str )
{
if( !str.compare("BACKPROP") )
return CvANN_MLP_TrainParams::BACKPROP;
return ANN_MLP::Params::BACKPROP;
if( !str.compare("RPROP") )
return CvANN_MLP_TrainParams::RPROP;
return ANN_MLP::Params::RPROP;
CV_Error( CV_StsBadArg, "incorrect ann train method string" );
return -1;
}
void ann_check_data_and_get_predictors( CvMLData* _data, CvMat* _inputs )
void ann_check_data( Ptr<TrainData> _data )
{
const CvMat* values = _data->get_values();
const CvMat* var_idx = _data->get_var_idx();
if( var_idx->cols + var_idx->rows != values->cols )
Mat values = _data->getSamples();
Mat var_idx = _data->getVarIdx();
int nvars = (int)var_idx.total();
if( nvars != 0 && nvars != values.cols )
CV_Error( CV_StsBadArg, "var_idx is not supported" );
if( _data->get_missing() )
if( !_data->getMissing().empty() )
CV_Error( CV_StsBadArg, "missing values are not supported" );
int resp_idx = _data->get_response_idx();
if( resp_idx == 0)
cvGetCols( values, _inputs, 1, values->cols );
else if( resp_idx == values->cols - 1 )
cvGetCols( values, _inputs, 0, values->cols - 1 );
else
CV_Error( CV_StsBadArg, "outputs must be in the first or last column; other cases are not supported" );
}
void ann_get_new_responses( CvMLData* _data, Mat& new_responses, map<int, int>& cls_map )
// unroll the categorical responses to binary vectors
Mat ann_get_new_responses( Ptr<TrainData> _data, map<int, int>& cls_map )
{
const CvMat* train_sidx = _data->get_train_sample_idx();
int* train_sidx_ptr = train_sidx->data.i;
const CvMat* responses = _data->get_responses();
float* responses_ptr = responses->data.fl;
int r_step = CV_IS_MAT_CONT(responses->type) ?
1 : responses->step / CV_ELEM_SIZE(responses->type);
Mat train_sidx = _data->getTrainSampleIdx();
int* train_sidx_ptr = train_sidx.ptr<int>();
Mat responses = _data->getResponses();
int cls_count = 0;
// construct cls_map
cls_map.clear();
for( int si = 0; si < train_sidx->cols; si++ )
int nresponses = (int)responses.total();
int si, n = !train_sidx.empty() ? (int)train_sidx.total() : nresponses;
for( si = 0; si < n; si++ )
{
int sidx = train_sidx_ptr[si];
int r = cvRound(responses_ptr[sidx*r_step]);
CV_DbgAssert( fabs(responses_ptr[sidx*r_step]-r) < FLT_EPSILON );
int cls_map_size = (int)cls_map.size();
cls_map[r];
if ( (int)cls_map.size() > cls_map_size )
int sidx = train_sidx_ptr ? train_sidx_ptr[si] : si;
int r = cvRound(responses.at<float>(sidx));
CV_DbgAssert( fabs(responses.at<float>(sidx) - r) < FLT_EPSILON );
map<int,int>::iterator it = cls_map.find(r);
if( it == cls_map.end() )
cls_map[r] = cls_count++;
}
new_responses.create( responses->rows, cls_count, CV_32F );
new_responses.setTo( 0 );
for( int si = 0; si < train_sidx->cols; si++ )
Mat new_responses = Mat::zeros( nresponses, cls_count, CV_32F );
for( si = 0; si < n; si++ )
{
int sidx = train_sidx_ptr[si];
int r = cvRound(responses_ptr[sidx*r_step]);
int sidx = train_sidx_ptr ? train_sidx_ptr[si] : si;
int r = cvRound(responses.at<float>(sidx));
int cidx = cls_map[r];
new_responses.ptr<float>(sidx)[cidx] = 1;
new_responses.at<float>(sidx, cidx) = 1.f;
}
return new_responses;
}
int ann_train( CvANN_MLP* ann, CvMLData* _data, Mat& new_responses, CvANN_MLP_TrainParams _params, int flags = 0 )
{
const CvMat* train_sidx = _data->get_train_sample_idx();
CvMat predictors;
ann_check_data_and_get_predictors( _data, &predictors );
CvMat _new_responses = CvMat( new_responses );
return ann->train( &predictors, &_new_responses, 0, train_sidx, _params, flags );
}
float ann_calc_error( CvANN_MLP* ann, CvMLData* _data, map<int, int>& cls_map, int type , vector<float> *resp_labels )
float ann_calc_error( Ptr<StatModel> ann, Ptr<TrainData> _data, map<int, int>& cls_map, int type, vector<float> *resp_labels )
{
float err = 0;
const CvMat* responses = _data->get_responses();
const CvMat* sample_idx = (type == CV_TEST_ERROR) ? _data->get_test_sample_idx() : _data->get_train_sample_idx();
int* sidx = sample_idx ? sample_idx->data.i : 0;
int r_step = CV_IS_MAT_CONT(responses->type) ?
1 : responses->step / CV_ELEM_SIZE(responses->type);
CvMat predictors;
ann_check_data_and_get_predictors( _data, &predictors );
int sample_count = sample_idx ? sample_idx->cols : 0;
sample_count = (type == CV_TRAIN_ERROR && sample_count == 0) ? predictors.rows : sample_count;
Mat samples = _data->getSamples();
Mat responses = _data->getResponses();
Mat sample_idx = (type == CV_TEST_ERROR) ? _data->getTestSampleIdx() : _data->getTrainSampleIdx();
int* sidx = !sample_idx.empty() ? sample_idx.ptr<int>() : 0;
ann_check_data( _data );
int sample_count = (int)sample_idx.total();
sample_count = (type == CV_TRAIN_ERROR && sample_count == 0) ? samples.rows : sample_count;
float* pred_resp = 0;
vector<float> innresp;
if( sample_count > 0 )
@@ -392,17 +171,16 @@ float ann_calc_error( CvANN_MLP* ann, CvMLData* _data, map<int, int>& cls_map, i
}
int cls_count = (int)cls_map.size();
Mat output( 1, cls_count, CV_32FC1 );
CvMat _output = CvMat(output);
for( int i = 0; i < sample_count; i++ )
{
CvMat sample;
int si = sidx ? sidx[i] : i;
cvGetRow( &predictors, &sample, si );
ann->predict( &sample, &_output );
CvPoint best_cls;
cvMinMaxLoc( &_output, 0, 0, 0, &best_cls, 0 );
int r = cvRound(responses->data.fl[si*r_step]);
CV_DbgAssert( fabs(responses->data.fl[si*r_step]-r) < FLT_EPSILON );
Mat sample = samples.row(si);
ann->predict( sample, output );
Point best_cls;
minMaxLoc(output, 0, 0, 0, &best_cls, 0);
int r = cvRound(responses.at<float>(si));
CV_DbgAssert( fabs(responses.at<float>(si) - r) < FLT_EPSILON );
r = cls_map[r];
int d = best_cls.x == r ? 0 : 1;
err += d;
@@ -417,13 +195,13 @@ float ann_calc_error( CvANN_MLP* ann, CvMLData* _data, map<int, int>& cls_map, i
int str_to_boost_type( String& str )
{
if ( !str.compare("DISCRETE") )
return CvBoost::DISCRETE;
return Boost::DISCRETE;
if ( !str.compare("REAL") )
return CvBoost::REAL;
return Boost::REAL;
if ( !str.compare("LOGIT") )
return CvBoost::LOGIT;
return Boost::LOGIT;
if ( !str.compare("GENTLE") )
return CvBoost::GENTLE;
return Boost::GENTLE;
CV_Error( CV_StsBadArg, "incorrect boost type string" );
return -1;
}
@@ -446,76 +224,37 @@ CV_MLBaseTest::CV_MLBaseTest(const char* _modelName)
RNG& rng = theRNG();
initSeed = rng.state;
rng.state = seeds[rng(seedCount)];
modelName = _modelName;
nbayes = 0;
knearest = 0;
svm = 0;
ann = 0;
dtree = 0;
boost = 0;
rtrees = 0;
ertrees = 0;
if( !modelName.compare(CV_NBAYES) )
nbayes = new CvNormalBayesClassifier;
else if( !modelName.compare(CV_KNEAREST) )
knearest = new CvKNearest;
else if( !modelName.compare(CV_SVM) )
svm = new CvSVM;
else if( !modelName.compare(CV_ANN) )
ann = new CvANN_MLP;
else if( !modelName.compare(CV_DTREE) )
dtree = new CvDTree;
else if( !modelName.compare(CV_BOOST) )
boost = new CvBoost;
else if( !modelName.compare(CV_RTREES) )
rtrees = new CvRTrees;
else if( !modelName.compare(CV_ERTREES) )
ertrees = new CvERTrees;
}
CV_MLBaseTest::~CV_MLBaseTest()
{
if( validationFS.isOpened() )
validationFS.release();
if( nbayes )
delete nbayes;
if( knearest )
delete knearest;
if( svm )
delete svm;
if( ann )
delete ann;
if( dtree )
delete dtree;
if( boost )
delete boost;
if( rtrees )
delete rtrees;
if( ertrees )
delete ertrees;
theRNG().state = initSeed;
}
int CV_MLBaseTest::read_params( CvFileStorage* _fs )
int CV_MLBaseTest::read_params( CvFileStorage* __fs )
{
if( !_fs )
FileStorage _fs(__fs, false);
if( !_fs.isOpened() )
test_case_count = -1;
else
{
CvFileNode* fn = cvGetRootFileNode( _fs, 0 );
fn = (CvFileNode*)cvGetSeqElem( fn->data.seq, 0 );
fn = cvGetFileNodeByName( _fs, fn, "run_params" );
CvSeq* dataSetNamesSeq = cvGetFileNodeByName( _fs, fn, modelName.c_str() )->data.seq;
test_case_count = dataSetNamesSeq ? dataSetNamesSeq->total : -1;
FileNode fn = _fs.getFirstTopLevelNode()["run_params"][modelName];
test_case_count = (int)fn.size();
if( test_case_count <= 0 )
test_case_count = -1;
if( test_case_count > 0 )
{
dataSetNames.resize( test_case_count );
vector<string>::iterator it = dataSetNames.begin();
for( int i = 0; i < test_case_count; i++, it++ )
*it = ((CvFileNode*)cvGetSeqElem( dataSetNamesSeq, i ))->data.str.ptr;
FileNodeIterator it = fn.begin();
for( int i = 0; i < test_case_count; i++, ++it )
{
dataSetNames[i] = (string)*it;
}
}
}
return cvtest::TS::OK;;
@@ -547,8 +286,6 @@ void CV_MLBaseTest::run( int )
int CV_MLBaseTest::prepare_test_case( int test_case_idx )
{
int trainSampleCount, respIdx;
String varTypes;
clear();
string dataPath = ts->get_data_path();
@@ -560,30 +297,27 @@ int CV_MLBaseTest::prepare_test_case( int test_case_idx )
string dataName = dataSetNames[test_case_idx],
filename = dataPath + dataName + ".data";
if ( data.read_csv( filename.c_str() ) != 0)
{
char msg[100];
sprintf( msg, "file %s can not be read", filename.c_str() );
ts->printf( cvtest::TS::LOG, msg );
return cvtest::TS::FAIL_INVALID_TEST_DATA;
}
FileNode dataParamsNode = validationFS.getFirstTopLevelNode()["validation"][modelName][dataName]["data_params"];
CV_DbgAssert( !dataParamsNode.empty() );
CV_DbgAssert( !dataParamsNode["LS"].empty() );
dataParamsNode["LS"] >> trainSampleCount;
CvTrainTestSplit spl( trainSampleCount );
data.set_train_test_split( &spl );
int trainSampleCount = (int)dataParamsNode["LS"];
CV_DbgAssert( !dataParamsNode["resp_idx"].empty() );
dataParamsNode["resp_idx"] >> respIdx;
data.set_response_idx( respIdx );
int respIdx = (int)dataParamsNode["resp_idx"];
CV_DbgAssert( !dataParamsNode["types"].empty() );
dataParamsNode["types"] >> varTypes;
data.set_var_types( varTypes.c_str() );
String varTypes = (String)dataParamsNode["types"];
data = TrainData::loadFromCSV(filename, 0, respIdx, respIdx+1, varTypes);
if( data.empty() )
{
ts->printf( cvtest::TS::LOG, "file %s can not be read\n", filename.c_str() );
return cvtest::TS::FAIL_INVALID_TEST_DATA;
}
data->setTrainTestSplit(trainSampleCount);
return cvtest::TS::OK;
}
@@ -598,114 +332,97 @@ int CV_MLBaseTest::train( int testCaseIdx )
FileNode modelParamsNode =
validationFS.getFirstTopLevelNode()["validation"][modelName][dataSetNames[testCaseIdx]]["model_params"];
if( !modelName.compare(CV_NBAYES) )
is_trained = nbayes_train( nbayes, &data );
else if( !modelName.compare(CV_KNEAREST) )
if( modelName == CV_NBAYES )
model = NormalBayesClassifier::create();
else if( modelName == CV_KNEAREST )
{
assert( 0 );
//is_trained = knearest->train( &data );
model = KNearest::create();
}
else if( !modelName.compare(CV_SVM) )
else if( modelName == CV_SVM )
{
String svm_type_str, kernel_type_str;
modelParamsNode["svm_type"] >> svm_type_str;
modelParamsNode["kernel_type"] >> kernel_type_str;
CvSVMParams params;
params.svm_type = str_to_svm_type( svm_type_str );
params.kernel_type = str_to_svm_kernel_type( kernel_type_str );
SVM::Params params;
params.svmType = str_to_svm_type( svm_type_str );
params.kernelType = str_to_svm_kernel_type( kernel_type_str );
modelParamsNode["degree"] >> params.degree;
modelParamsNode["gamma"] >> params.gamma;
modelParamsNode["coef0"] >> params.coef0;
modelParamsNode["C"] >> params.C;
modelParamsNode["nu"] >> params.nu;
modelParamsNode["p"] >> params.p;
is_trained = svm_train( svm, &data, params );
model = SVM::create(params);
}
else if( !modelName.compare(CV_EM) )
else if( modelName == CV_EM )
{
assert( 0 );
}
else if( !modelName.compare(CV_ANN) )
else if( modelName == CV_ANN )
{
String train_method_str;
double param1, param2;
modelParamsNode["train_method"] >> train_method_str;
modelParamsNode["param1"] >> param1;
modelParamsNode["param2"] >> param2;
Mat new_responses;
ann_get_new_responses( &data, new_responses, cls_map );
int layer_sz[] = { data.get_values()->cols - 1, 100, 100, (int)cls_map.size() };
CvMat layer_sizes =
cvMat( 1, (int)(sizeof(layer_sz)/sizeof(layer_sz[0])), CV_32S, layer_sz );
ann->create( &layer_sizes );
is_trained = ann_train( ann, &data, new_responses, CvANN_MLP_TrainParams(cvTermCriteria(CV_TERMCRIT_ITER,300,0.01),
str_to_ann_train_method(train_method_str), param1, param2) ) >= 0;
Mat new_responses = ann_get_new_responses( data, cls_map );
// binarize the responses
data = TrainData::create(data->getSamples(), data->getLayout(), new_responses,
data->getVarIdx(), data->getTrainSampleIdx());
int layer_sz[] = { data->getNAllVars(), 100, 100, (int)cls_map.size() };
Mat layer_sizes( 1, (int)(sizeof(layer_sz)/sizeof(layer_sz[0])), CV_32S, layer_sz );
model = ANN_MLP::create(layer_sizes, ANN_MLP::Params(TermCriteria(TermCriteria::COUNT,300,0.01),
str_to_ann_train_method(train_method_str), param1, param2));
}
else if( !modelName.compare(CV_DTREE) )
else if( modelName == CV_DTREE )
{
int MAX_DEPTH, MIN_SAMPLE_COUNT, MAX_CATEGORIES, CV_FOLDS;
float REG_ACCURACY = 0;
bool USE_SURROGATE, IS_PRUNED;
bool USE_SURROGATE = false, IS_PRUNED;
modelParamsNode["max_depth"] >> MAX_DEPTH;
modelParamsNode["min_sample_count"] >> MIN_SAMPLE_COUNT;
modelParamsNode["use_surrogate"] >> USE_SURROGATE;
//modelParamsNode["use_surrogate"] >> USE_SURROGATE;
modelParamsNode["max_categories"] >> MAX_CATEGORIES;
modelParamsNode["cv_folds"] >> CV_FOLDS;
modelParamsNode["is_pruned"] >> IS_PRUNED;
is_trained = dtree->train( &data,
CvDTreeParams(MAX_DEPTH, MIN_SAMPLE_COUNT, REG_ACCURACY, USE_SURROGATE,
MAX_CATEGORIES, CV_FOLDS, false, IS_PRUNED, 0 )) != 0;
model = DTrees::create(DTrees::Params(MAX_DEPTH, MIN_SAMPLE_COUNT, REG_ACCURACY, USE_SURROGATE,
MAX_CATEGORIES, CV_FOLDS, false, IS_PRUNED, Mat() ));
}
else if( !modelName.compare(CV_BOOST) )
else if( modelName == CV_BOOST )
{
int BOOST_TYPE, WEAK_COUNT, MAX_DEPTH;
float WEIGHT_TRIM_RATE;
bool USE_SURROGATE;
bool USE_SURROGATE = false;
String typeStr;
modelParamsNode["type"] >> typeStr;
BOOST_TYPE = str_to_boost_type( typeStr );
modelParamsNode["weak_count"] >> WEAK_COUNT;
modelParamsNode["weight_trim_rate"] >> WEIGHT_TRIM_RATE;
modelParamsNode["max_depth"] >> MAX_DEPTH;
modelParamsNode["use_surrogate"] >> USE_SURROGATE;
is_trained = boost->train( &data,
CvBoostParams(BOOST_TYPE, WEAK_COUNT, WEIGHT_TRIM_RATE, MAX_DEPTH, USE_SURROGATE, 0) ) != 0;
//modelParamsNode["use_surrogate"] >> USE_SURROGATE;
model = Boost::create( Boost::Params(BOOST_TYPE, WEAK_COUNT, WEIGHT_TRIM_RATE, MAX_DEPTH, USE_SURROGATE, Mat()) );
}
else if( !modelName.compare(CV_RTREES) )
else if( modelName == CV_RTREES )
{
int MAX_DEPTH, MIN_SAMPLE_COUNT, MAX_CATEGORIES, CV_FOLDS, NACTIVE_VARS, MAX_TREES_NUM;
float REG_ACCURACY = 0, OOB_EPS = 0.0;
bool USE_SURROGATE, IS_PRUNED;
bool USE_SURROGATE = false, IS_PRUNED;
modelParamsNode["max_depth"] >> MAX_DEPTH;
modelParamsNode["min_sample_count"] >> MIN_SAMPLE_COUNT;
modelParamsNode["use_surrogate"] >> USE_SURROGATE;
//modelParamsNode["use_surrogate"] >> USE_SURROGATE;
modelParamsNode["max_categories"] >> MAX_CATEGORIES;
modelParamsNode["cv_folds"] >> CV_FOLDS;
modelParamsNode["is_pruned"] >> IS_PRUNED;
modelParamsNode["nactive_vars"] >> NACTIVE_VARS;
modelParamsNode["max_trees_num"] >> MAX_TREES_NUM;
is_trained = rtrees->train( &data, CvRTParams( MAX_DEPTH, MIN_SAMPLE_COUNT, REG_ACCURACY,
USE_SURROGATE, MAX_CATEGORIES, 0, true, // (calc_var_importance == true) <=> RF processes variable importance
NACTIVE_VARS, MAX_TREES_NUM, OOB_EPS, CV_TERMCRIT_ITER)) != 0;
}
else if( !modelName.compare(CV_ERTREES) )
{
int MAX_DEPTH, MIN_SAMPLE_COUNT, MAX_CATEGORIES, CV_FOLDS, NACTIVE_VARS, MAX_TREES_NUM;
float REG_ACCURACY = 0, OOB_EPS = 0.0;
bool USE_SURROGATE, IS_PRUNED;
modelParamsNode["max_depth"] >> MAX_DEPTH;
modelParamsNode["min_sample_count"] >> MIN_SAMPLE_COUNT;
modelParamsNode["use_surrogate"] >> USE_SURROGATE;
modelParamsNode["max_categories"] >> MAX_CATEGORIES;
modelParamsNode["cv_folds"] >> CV_FOLDS;
modelParamsNode["is_pruned"] >> IS_PRUNED;
modelParamsNode["nactive_vars"] >> NACTIVE_VARS;
modelParamsNode["max_trees_num"] >> MAX_TREES_NUM;
is_trained = ertrees->train( &data, CvRTParams( MAX_DEPTH, MIN_SAMPLE_COUNT, REG_ACCURACY,
USE_SURROGATE, MAX_CATEGORIES, 0, false, // (calc_var_importance == true) <=> RF processes variable importance
NACTIVE_VARS, MAX_TREES_NUM, OOB_EPS, CV_TERMCRIT_ITER)) != 0;
model = RTrees::create(RTrees::Params( MAX_DEPTH, MIN_SAMPLE_COUNT, REG_ACCURACY,
USE_SURROGATE, MAX_CATEGORIES, Mat(), true, // (calc_var_importance == true) <=> RF processes variable importance
NACTIVE_VARS, TermCriteria(TermCriteria::COUNT, MAX_TREES_NUM, OOB_EPS)));
}
if( !model.empty() )
is_trained = model->train(data, 0);
if( !is_trained )
{
ts->printf( cvtest::TS::LOG, "in test case %d model training was failed", testCaseIdx );
@@ -714,78 +431,46 @@ int CV_MLBaseTest::train( int testCaseIdx )
return cvtest::TS::OK;
}
float CV_MLBaseTest::get_error( int /*testCaseIdx*/, int type, vector<float> *resp )
float CV_MLBaseTest::get_test_error( int /*testCaseIdx*/, vector<float> *resp )
{
int type = CV_TEST_ERROR;
float err = 0;
if( !modelName.compare(CV_NBAYES) )
err = nbayes_calc_error( nbayes, &data, type, resp );
else if( !modelName.compare(CV_KNEAREST) )
{
Mat _resp;
if( modelName == CV_EM )
assert( 0 );
/*testCaseIdx = 0;
int k = 2;
validationFS.getFirstTopLevelNode()["validation"][modelName][dataSetNames[testCaseIdx]]["model_params"]["k"] >> k;
err = knearest->calc_error( &data, k, type, resp );*/
}
else if( !modelName.compare(CV_SVM) )
err = svm_calc_error( svm, &data, type, resp );
else if( !modelName.compare(CV_EM) )
assert( 0 );
else if( !modelName.compare(CV_ANN) )
err = ann_calc_error( ann, &data, cls_map, type, resp );
else if( !modelName.compare(CV_DTREE) )
err = dtree->calc_error( &data, type, resp );
else if( !modelName.compare(CV_BOOST) )
err = boost->calc_error( &data, type, resp );
else if( !modelName.compare(CV_RTREES) )
err = rtrees->calc_error( &data, type, resp );
else if( !modelName.compare(CV_ERTREES) )
err = ertrees->calc_error( &data, type, resp );
else if( modelName == CV_ANN )
err = ann_calc_error( model, data, cls_map, type, resp );
else if( modelName == CV_DTREE || modelName == CV_BOOST || modelName == CV_RTREES ||
modelName == CV_SVM || modelName == CV_NBAYES || modelName == CV_KNEAREST )
err = model->calcError( data, true, _resp );
if( !_resp.empty() && resp )
_resp.convertTo(*resp, CV_32F);
return err;
}
void CV_MLBaseTest::save( const char* filename )
{
if( !modelName.compare(CV_NBAYES) )
nbayes->save( filename );
else if( !modelName.compare(CV_KNEAREST) )
knearest->save( filename );
else if( !modelName.compare(CV_SVM) )
svm->save( filename );
else if( !modelName.compare(CV_ANN) )
ann->save( filename );
else if( !modelName.compare(CV_DTREE) )
dtree->save( filename );
else if( !modelName.compare(CV_BOOST) )
boost->save( filename );
else if( !modelName.compare(CV_RTREES) )
rtrees->save( filename );
else if( !modelName.compare(CV_ERTREES) )
ertrees->save( filename );
model->save( filename );
}
void CV_MLBaseTest::load( const char* filename )
{
if( !modelName.compare(CV_NBAYES) )
nbayes->load( filename );
else if( !modelName.compare(CV_KNEAREST) )
knearest->load( filename );
else if( !modelName.compare(CV_SVM) )
{
delete svm;
svm = new CvSVM;
svm->load( filename );
}
else if( !modelName.compare(CV_ANN) )
ann->load( filename );
else if( !modelName.compare(CV_DTREE) )
dtree->load( filename );
else if( !modelName.compare(CV_BOOST) )
boost->load( filename );
else if( !modelName.compare(CV_RTREES) )
rtrees->load( filename );
else if( !modelName.compare(CV_ERTREES) )
ertrees->load( filename );
if( modelName == CV_NBAYES )
model = StatModel::load<NormalBayesClassifier>( filename );
else if( modelName == CV_KNEAREST )
model = StatModel::load<KNearest>( filename );
else if( modelName == CV_SVM )
model = StatModel::load<SVM>( filename );
else if( modelName == CV_ANN )
model = StatModel::load<ANN_MLP>( filename );
else if( modelName == CV_DTREE )
model = StatModel::load<DTrees>( filename );
else if( modelName == CV_BOOST )
model = StatModel::load<Boost>( filename );
else if( modelName == CV_RTREES )
model = StatModel::load<RTrees>( filename );
else
CV_Error( CV_StsNotImplemented, "invalid stat model name");
}
/* End of file. */