Skip to content

O
opencv
Commit 10b60f8d authored by Vadim Pisarevsky's avatar Vadim Pisarevsky
Browse files
Options

continuing refactoring ml samples; added "max vote" response to ANN_MLP.… 

continuing refactoring ml samples; added "max vote" response to ANN_MLP. Probably, should make it in less hacky way
parent 223cdcd0
Hide whitespace changes
Inline Side-by-side
Showing with 273 additions and 353 deletions
modules/ml/src/ann_mlp.cpp
View file @ 10b60f8d
...@@ -228,9 +228,8 @@ public: ...@@ -228,9 +228,8 @@ public:
int n = inputs.rows, dn0 = n; int n = inputs.rows, dn0 = n;
CV_Assert( (type == CV_32F || type == CV_64F) && inputs.cols == layer_sizes[0] ); CV_Assert( (type == CV_32F || type == CV_64F) && inputs.cols == layer_sizes[0] );
_outputs.create(n, layer_sizes[l_count-1], type); int noutputs = layer_sizes[l_count-1];
Mat outputs;
Mat outputs = _outputs.getMat();
int min_buf_sz = 2*max_lsize; int min_buf_sz = 2*max_lsize;
int buf_sz = n*min_buf_sz; int buf_sz = n*min_buf_sz;
...@@ -242,9 +241,20 @@ public: ...@@ -242,9 +241,20 @@ public:
buf_sz = dn0*min_buf_sz; buf_sz = dn0*min_buf_sz;
} }
cv::AutoBuffer<double> _buf(buf_sz); cv::AutoBuffer<double> _buf(buf_sz+noutputs);
double* buf = _buf; double* buf = _buf;
if( !_outputs.needed() )
{
CV_Assert( n == 1 );
outputs = Mat(n, noutputs, type, buf + buf_sz);
}
else
{
_outputs.create(n, noutputs, type);
outputs = _outputs.getMat();
}
int dn = 0; int dn = 0;
for( int i = 0; i < n; i += dn ) for( int i = 0; i < n; i += dn )
{ {
...@@ -273,6 +283,13 @@ public: ...@@ -273,6 +283,13 @@ public:
scale_output( layer_in, layer_out ); scale_output( layer_in, layer_out );
} }
if( n == 1 )
{
int maxIdx[] = {0, 0};
minMaxIdx(outputs, 0, 0, 0, maxIdx);
return maxIdx[0] + maxIdx[1];
}
return 0.f; return 0.f;
} }
......
samples/cpp/letter_recog.cpp
View file @ 10b60f8d
#include "opencv2/core/core_c.h" #include "opencv2/core/core.hpp"
#include "opencv2/ml/ml.hpp" #include "opencv2/ml/ml.hpp"
#include <cstdio> #include <cstdio>
#include <vector> #include <vector>
/* #include <iostream>
*/ using namespace std;
using namespace cv;
using namespace cv::ml;
static void help() static void help()
{ {
...@@ -33,142 +35,101 @@ static void help() ...@@ -33,142 +35,101 @@ static void help()
} }
// This function reads data and responses from the file <filename> // This function reads data and responses from the file <filename>
static int static bool
read_num_class_data( const char* filename, int var_count, read_num_class_data( const string& filename, int var_count,
CvMat** data, CvMat** responses ) Mat* _data, Mat* _responses )
{ {
const int M = 1024; const int M = 1024;
FILE* f = fopen( filename, "rt" );
CvMemStorage* storage;
CvSeq* seq;
char buf[M+2]; char buf[M+2];
float* el_ptr;
CvSeqReader reader;
int i, j;
if( !f ) Mat el_ptr(1, var_count, CV_32F);
return 0; int i;
vector<int> responses;
el_ptr = new float[var_count+1]; _data->release();
storage = cvCreateMemStorage(); _responses->release();
seq = cvCreateSeq( 0, sizeof(*seq), (var_count+1)*sizeof(float), storage );
FILE* f = fopen( filename.c_str(), "rt" );
if( !f )
{
cout << "Could not read the database " << filename << endl;
return false;
}
for(;;) for(;;)
{ {
char* ptr; char* ptr;
if( !fgets( buf, M, f ) || !strchr( buf, ',' ) ) if( !fgets( buf, M, f ) || !strchr( buf, ',' ) )
break; break;
el_ptr[0] = buf[0]; responses.push_back((int)buf[0]);
ptr = buf+2; ptr = buf+2;
for( i = 1; i <= var_count; i++ ) for( i = 0; i < var_count; i++ )
{ {
int n = 0; int n = 0;
sscanf( ptr, "%f%n", el_ptr + i, &n ); sscanf( ptr, "%f%n", &el_ptr.at<float>(i), &n );
ptr += n + 1; ptr += n + 1;
} }
if( i <= var_count ) if( i < var_count )
break; break;
cvSeqPush( seq, el_ptr ); _data->push_back(el_ptr);
} }
fclose(f); fclose(f);
Mat(responses).copyTo(*_responses);
*data = cvCreateMat( seq->total, var_count, CV_32F ); cout << "The database " << filename << " is loaded.\n";
*responses = cvCreateMat( seq->total, 1, CV_32F );
cvStartReadSeq( seq, &reader );
for( i = 0; i < seq->total; i++ )
{
const float* sdata = (float*)reader.ptr + 1;
float* ddata = data[0]->data.fl + var_count*i;
float* dr = responses[0]->data.fl + i;
for( j = 0; j < var_count; j++ )
ddata[j] = sdata[j];
*dr = sdata[-1];
CV_NEXT_SEQ_ELEM( seq->elem_size, reader );
}
cvReleaseMemStorage( &storage ); return true;
delete[] el_ptr;
return 1;
} }
static template<typename T>
int build_rtrees_classifier( char* data_filename, static Ptr<T> load_classifier(const string& filename_to_load)
char* filename_to_save, char* filename_to_load )
{ {
CvMat* data = 0; // load classifier from the specified file
CvMat* responses = 0; Ptr<T> model = StatModel::load<T>( filename_to_load );
CvMat* var_type = 0; if( model.empty() )
CvMat* sample_idx = 0; cout << "Could not read the classifier " << filename_to_load << endl;
else
int ok = read_num_class_data( data_filename, 16, &data, &responses ); cout << "The classifier " << filename_to_load << " is loaded.\n";
int nsamples_all = 0, ntrain_samples = 0;
int i = 0;
double train_hr = 0, test_hr = 0;
CvRTrees forest;
CvMat* var_importance = 0;
if( !ok )
{
printf( "Could not read the database %s\n", data_filename );
return -1;
}
printf( "The database %s is loaded.\n", data_filename ); return model;
nsamples_all = data->rows; }
ntrain_samples = (int)(nsamples_all*0.8);
// Create or load Random Trees classifier static Ptr<TrainData>
if( filename_to_load ) prepare_train_data(const Mat& data, const Mat& responses, int ntrain_samples)
{ {
// load classifier from the specified file Mat sample_idx = Mat::zeros( 1, data.rows, CV_8U );
forest.load( filename_to_load ); Mat train_samples = sample_idx.colRange(0, ntrain_samples);
ntrain_samples = 0; train_samples.setTo(Scalar::all(1));
if( forest.get_tree_count() == 0 )
{
printf( "Could not read the classifier %s\n", filename_to_load );
return -1;
}
printf( "The classifier %s is loaded.\n", filename_to_load );
}
else
{
// create classifier by using <data> and <responses>
printf( "Training the classifier ...\n");
// 1. create type mask int nvars = data.cols;
var_type = cvCreateMat( data->cols + 1, 1, CV_8U ); Mat var_type( nvars + 1, 1, CV_8U );
cvSet( var_type, cvScalarAll(CV_VAR_ORDERED) ); var_type.setTo(Scalar::all(VAR_ORDERED));
cvSetReal1D( var_type, data->cols, CV_VAR_CATEGORICAL ); var_type.at<uchar>(nvars) = VAR_CATEGORICAL;
// 2. create sample_idx return TrainData::create(data, ROW_SAMPLE, responses,
sample_idx = cvCreateMat( 1, nsamples_all, CV_8UC1 ); noArray(), sample_idx, noArray(), var_type);
{ }
CvMat mat;
cvGetCols( sample_idx, &mat, 0, ntrain_samples );
cvSet( &mat, cvRealScalar(1) );
cvGetCols( sample_idx, &mat, ntrain_samples, nsamples_all ); inline TermCriteria TC(int iters, double eps)
cvSetZero( &mat ); {
} return TermCriteria(TermCriteria::MAX_ITER + (eps > 0 ? TermCriteria::EPS : 0), iters, eps);
}
// 3. train classifier static void test_and_save_classifier(const Ptr<StatModel>& model,
forest.train( data, CV_ROW_SAMPLE, responses, 0, sample_idx, var_type, 0, const Mat& data, const Mat& responses,
CvRTParams(10,10,0,false,15,0,true,4,100,0.01f,CV_TERMCRIT_ITER)); int ntrain_samples, int rdelta,
printf( "\n"); const string& filename_to_save)
} {
int i, nsamples_all = data.rows;
double train_hr = 0, test_hr = 0;
// compute prediction error on train and test data // compute prediction error on train and test data
for( i = 0; i < nsamples_all; i++ ) for( i = 0; i < nsamples_all; i++ )
{ {
double r; Mat sample = data.row(i);
CvMat sample;
cvGetRow( data, &sample, i );
r = forest.predict( &sample ); float r = model->predict( sample );
r = fabs((double)r - responses->data.fl[i]) <= FLT_EPSILON ? 1 : 0; r = std::abs(r + rdelta - responses.at<int>(i)) <= FLT_EPSILON ? 1 : 0;
if( i < ntrain_samples ) if( i < ntrain_samples )
train_hr += r; train_hr += r;
...@@ -176,93 +137,101 @@ int build_rtrees_classifier( char* data_filename, ...@@ -176,93 +137,101 @@ int build_rtrees_classifier( char* data_filename,
test_hr += r; test_hr += r;
} }
test_hr /= (double)(nsamples_all-ntrain_samples); test_hr /= nsamples_all - ntrain_samples;
train_hr /= (double)ntrain_samples; train_hr = ntrain_samples > 0 ? train_hr/ntrain_samples : 1.;
printf( "Recognition rate: train = %.1f%%, test = %.1f%%\n", printf( "Recognition rate: train = %.1f%%, test = %.1f%%\n",
train_hr*100., test_hr*100. ); train_hr*100., test_hr*100. );
printf( "Number of trees: %d\n", forest.get_tree_count() ); if( !filename_to_save.empty() )
// Print variable importance
var_importance = (CvMat*)forest.get_var_importance();
if( var_importance )
{ {
double rt_imp_sum = cvSum( var_importance ).val[0]; model->save( filename_to_save );
printf("var#\timportance (in %%):\n");
for( i = 0; i < var_importance->cols; i++ )
printf( "%-2d\t%-4.1f\n", i,
100.f*var_importance->data.fl[i]/rt_imp_sum);
} }
}
//Print some proximitites static bool
printf( "Proximities between some samples corresponding to the letter 'T':\n" ); build_rtrees_classifier( const string& data_filename,
const string& filename_to_save,
const string& filename_to_load )
{
Mat data;
Mat responses;
bool ok = read_num_class_data( data_filename, 16, &data, &responses );
if( !ok )
return ok;
Ptr<RTrees> model;
int nsamples_all = data.rows;
int ntrain_samples = (int)(nsamples_all*0.8);
// Create or load Random Trees classifier
if( !filename_to_load.empty() )
{ {
CvMat sample1, sample2; model = load_classifier<RTrees>(filename_to_load);
const int pairs[][2] = {{0,103}, {0,106}, {106,103}, {-1,-1}}; if( model.empty() )
return false;
ntrain_samples = 0;
}
else
{
// create classifier by using <data> and <responses>
cout << "Training the classifier ...\n";
Ptr<TrainData> tdata = prepare_train_data(data, responses, ntrain_samples);
for( i = 0; pairs[i][0] >= 0; i++ ) // 3. train classifier
{ model = RTrees::create(RTrees::Params(10,10,0,false,15,Mat(),true,4,TC(100,0.01f)));
cvGetRow( data, &sample1, pairs[i][0] ); model->train( tdata );
cvGetRow( data, &sample2, pairs[i][1] ); cout << endl;
printf( "proximity(%d,%d) = %.1f%%\n", pairs[i][0], pairs[i][1],
forest.get_proximity( &sample1, &sample2 )*100. );
}
} }
// Save Random Trees classifier to file if needed test_and_save_classifier(model, data, responses, ntrain_samples, 0, filename_to_save);
if( filename_to_save ) cout << "Number of trees: " << model->getRoots().size() << endl;
forest.save( filename_to_save );
cvReleaseMat( &sample_idx ); // Print variable importance
cvReleaseMat( &var_type ); Mat var_importance = model->getVarImportance();
cvReleaseMat( &data ); if( !var_importance.empty() )
cvReleaseMat( &responses ); {
double rt_imp_sum = sum( var_importance )[0];
printf("var#\timportance (in %%):\n");
int i, n = (int)var_importance.total();
for( i = 0; i < n; i++ )
printf( "%-2d\t%-4.1f\n", i, 100.f*var_importance.at<float>(i)/rt_imp_sum);
}
return 0; return true;
} }
static static bool
int build_boost_classifier( char* data_filename, build_boost_classifier( const string& data_filename,
char* filename_to_save, char* filename_to_load ) const string& filename_to_save,
const string& filename_to_load )
{ {
const int class_count = 26; const int class_count = 26;
CvMat* data = 0; Mat data;
CvMat* responses = 0; Mat responses;
CvMat* var_type = 0; Mat weak_responses;
CvMat* temp_sample = 0;
CvMat* weak_responses = 0;
int ok = read_num_class_data( data_filename, 16, &data, &responses );
int nsamples_all = 0, ntrain_samples = 0;
int var_count;
int i, j, k;
double train_hr = 0, test_hr = 0;
CvBoost boost;
bool ok = read_num_class_data( data_filename, 16, &data, &responses );
if( !ok ) if( !ok )
{ return ok;
printf( "Could not read the database %s\n", data_filename );
return -1;
}
printf( "The database %s is loaded.\n", data_filename ); int i, j, k;
nsamples_all = data->rows; Ptr<Boost> model;
ntrain_samples = (int)(nsamples_all*0.5);
var_count = data->cols; int nsamples_all = data.rows;
int ntrain_samples = (int)(nsamples_all*0.5);
int var_count = data.cols;
// Create or load Boosted Tree classifier // Create or load Boosted Tree classifier
if( filename_to_load ) if( !filename_to_load.empty() )
{ {
// load classifier from the specified file model = load_classifier<Boost>(filename_to_load);
boost.load( filename_to_load ); if( model.empty() )
return false;
ntrain_samples = 0; ntrain_samples = 0;
if( !boost.get_weak_predictors() )
{
printf( "Could not read the classifier %s\n", filename_to_load );
return -1;
}
printf( "The classifier %s is loaded.\n", filename_to_load );
} }
else else
{ {
...@@ -275,135 +244,108 @@ int build_boost_classifier( char* data_filename, ...@@ -275,135 +244,108 @@ int build_boost_classifier( char* data_filename,
// //
// !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! // !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
CvMat* new_data = cvCreateMat( ntrain_samples*class_count, var_count + 1, CV_32F ); Mat new_data( ntrain_samples*class_count, var_count + 1, CV_32F );
CvMat* new_responses = cvCreateMat( ntrain_samples*class_count, 1, CV_32S ); Mat new_responses( ntrain_samples*class_count, 1, CV_32S );
// 1. unroll the database type mask // 1. unroll the database type mask
printf( "Unrolling the database...\n"); printf( "Unrolling the database...\n");
for( i = 0; i < ntrain_samples; i++ ) for( i = 0; i < ntrain_samples; i++ )
{ {
float* data_row = (float*)(data->data.ptr + data->step*i); const float* data_row = data.ptr<float>(i);
for( j = 0; j < class_count; j++ ) for( j = 0; j < class_count; j++ )
{ {
float* new_data_row = (float*)(new_data->data.ptr + float* new_data_row = (float*)new_data.ptr<float>(i*class_count+j);
new_data->step*(i*class_count+j)); memcpy(new_data_row, data_row, var_count*sizeof(data_row[0]));
for( k = 0; k < var_count; k++ )
new_data_row[k] = data_row[k];
new_data_row[var_count] = (float)j; new_data_row[var_count] = (float)j;
new_responses->data.i[i*class_count + j] = responses->data.fl[i] == j+'A'; new_responses.at<int>(i*class_count + j) = responses.at<int>(i) == j+'A';
} }
} }
// 2. create type mask Mat var_type( 1, var_count + 2, CV_8U );
var_type = cvCreateMat( var_count + 2, 1, CV_8U ); var_type.setTo(Scalar::all(VAR_ORDERED));
cvSet( var_type, cvScalarAll(CV_VAR_ORDERED) ); var_type.at<uchar>(var_count) = var_type.at<uchar>(var_count+1) = VAR_CATEGORICAL;
// the last indicator variable, as well
// as the new (binary) response are categorical
cvSetReal1D( var_type, var_count, CV_VAR_CATEGORICAL );
cvSetReal1D( var_type, var_count+1, CV_VAR_CATEGORICAL );
// 3. train classifier Ptr<TrainData> tdata = TrainData::create(new_data, ROW_SAMPLE, new_responses,
printf( "Training the classifier (may take a few minutes)...\n"); noArray(), noArray(), noArray(), var_type);
boost.train( new_data, CV_ROW_SAMPLE, new_responses, 0, 0, var_type, 0, model = Boost::create(Boost::Params(Boost::REAL, 100, 0.95, 5, false, Mat() ));
CvBoostParams(CvBoost::REAL, 100, 0.95, 5, false, 0 ));
cvReleaseMat( &new_data ); cout << "Training the classifier (may take a few minutes)...\n";
cvReleaseMat( &new_responses ); model->train(tdata);
printf("\n"); cout << endl;
} }
temp_sample = cvCreateMat( 1, var_count + 1, CV_32F ); Mat temp_sample( 1, var_count + 1, CV_32F );
weak_responses = cvCreateMat( 1, boost.get_weak_predictors()->total, CV_32F ); float* tptr = temp_sample.ptr<float>();
// compute prediction error on train and test data // compute prediction error on train and test data
double train_hr = 0, test_hr = 0;
for( i = 0; i < nsamples_all; i++ ) for( i = 0; i < nsamples_all; i++ )
{ {
int best_class = 0; int best_class = 0;
double max_sum = -DBL_MAX; double max_sum = -DBL_MAX;
double r; const float* ptr = data.ptr<float>(i);
CvMat sample;
cvGetRow( data, &sample, i );
for( k = 0; k < var_count; k++ ) for( k = 0; k < var_count; k++ )
temp_sample->data.fl[k] = sample.data.fl[k]; tptr[k] = ptr[k];
for( j = 0; j < class_count; j++ ) for( j = 0; j < class_count; j++ )
{ {
temp_sample->data.fl[var_count] = (float)j; tptr[var_count] = (float)j;
boost.predict( temp_sample, 0, weak_responses ); float s = model->predict( temp_sample, noArray(), StatModel::RAW_OUTPUT );
double sum = cvSum( weak_responses ).val[0]; if( max_sum < s )
if( max_sum < sum )
{ {
max_sum = sum; max_sum = s;
best_class = j + 'A'; best_class = j + 'A';
} }
} }
r = fabs(best_class - responses->data.fl[i]) < FLT_EPSILON ? 1 : 0; double r = std::abs(best_class - responses.at<int>(i)) < FLT_EPSILON ? 1 : 0;
if( i < ntrain_samples ) if( i < ntrain_samples )
train_hr += r; train_hr += r;
else else
test_hr += r; test_hr += r;
} }
test_hr /= (double)(nsamples_all-ntrain_samples); test_hr /= nsamples_all-ntrain_samples;
train_hr /= (double)ntrain_samples; train_hr = ntrain_samples > 0 ? train_hr/ntrain_samples : 1.;
printf( "Recognition rate: train = %.1f%%, test = %.1f%%\n", printf( "Recognition rate: train = %.1f%%, test = %.1f%%\n",
train_hr*100., test_hr*100. ); train_hr*100., test_hr*100. );
printf( "Number of trees: %d\n", boost.get_weak_predictors()->total ); cout << "Number of trees: " << model->getRoots().size() << endl;
// Save classifier to file if needed // Save classifier to file if needed
if( filename_to_save ) if( !filename_to_save.empty() )
boost.save( filename_to_save ); model->save( filename_to_save );
cvReleaseMat( &temp_sample ); return true;
cvReleaseMat( &weak_responses );
cvReleaseMat( &var_type );
cvReleaseMat( &data );
cvReleaseMat( &responses );
return 0;
} }
static static bool
int build_mlp_classifier( char* data_filename, build_mlp_classifier( const string& data_filename,
char* filename_to_save, char* filename_to_load ) const string& filename_to_save,
const string& filename_to_load )
{ {
const int class_count = 26; const int class_count = 26;
CvMat* data = 0; Mat data;
CvMat train_data; Mat responses;
CvMat* responses = 0;
CvMat* mlp_response = 0;
int ok = read_num_class_data( data_filename, 16, &data, &responses );
int nsamples_all = 0, ntrain_samples = 0;
int i, j;
double train_hr = 0, test_hr = 0;
CvANN_MLP mlp;
bool ok = read_num_class_data( data_filename, 16, &data, &responses );
if( !ok ) if( !ok )
{ return ok;
printf( "Could not read the database %s\n", data_filename );
return -1;
}
printf( "The database %s is loaded.\n", data_filename ); int i, j;
nsamples_all = data->rows; Ptr<ANN_MLP> model;
ntrain_samples = (int)(nsamples_all*0.8);
int nsamples_all = data.rows;
int ntrain_samples = (int)(nsamples_all*0.8);
// Create or load MLP classifier // Create or load MLP classifier
if( filename_to_load ) if( !filename_to_load.empty() )
{ {
// load classifier from the specified file model = load_classifier<ANN_MLP>(filename_to_load);
mlp.load( filename_to_load ); if( model.empty() )
return false;
ntrain_samples = 0; ntrain_samples = 0;
if( !mlp.get_layer_count() )
{
printf( "Could not read the classifier %s\n", filename_to_load );
return -1;
}
printf( "The classifier %s is loaded.\n", filename_to_load );
} }
else else
{ {
...@@ -417,45 +359,44 @@ int build_mlp_classifier( char* data_filename, ...@@ -417,45 +359,44 @@ int build_mlp_classifier( char* data_filename,
// //
// !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! // !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
CvMat* new_responses = cvCreateMat( ntrain_samples, class_count, CV_32F ); Mat train_data = data.rowRange(0, ntrain_samples);
Mat new_responses = Mat::zeros( ntrain_samples, class_count, CV_32F );
// 1. unroll the responses // 1. unroll the responses
printf( "Unrolling the responses...\n"); cout << "Unrolling the responses...\n";
for( i = 0; i < ntrain_samples; i++ ) for( i = 0; i < ntrain_samples; i++ )
{ {
int cls_label = cvRound(responses->data.fl[i]) - 'A'; int cls_label = responses.at<int>(i) - 'A'
float* bit_vec = (float*)(new_responses->data.ptr + i*new_responses->step); new_responses.at<float>(i, cls_label) = 1.f;
for( j = 0; j < class_count; j++ )
bit_vec[j] = 0.f;
bit_vec[cls_label] = 1.f;
} }
cvGetRows( data, &train_data, 0, ntrain_samples );
// 2. train classifier // 2. train classifier
int layer_sz[] = { data->cols, 100, 100, class_count }; int layer_sz[] = { data.cols, 100, 100, class_count };
CvMat layer_sizes = int nlayers = (int)(sizeof(layer_sz)/sizeof(layer_sz[0]));
cvMat( 1, (int)(sizeof(layer_sz)/sizeof(layer_sz[0])), CV_32S, layer_sz ); Mat layer_sizes( 1, nlayers, CV_32S, layer_sz );
mlp.create( &layer_sizes );
printf( "Training the classifier (may take a few minutes)...\n");
#if 1 #if 1
int method = CvANN_MLP_TrainParams::BACKPROP; int method = ANN_MLP::Params::BACKPROP;
double method_param = 0.001; double method_param = 0.001;
int max_iter = 300; int max_iter = 300;
#else #else
int method = CvANN_MLP_TrainParams::RPROP; int method = ANN_MLP::Params::RPROP;
double method_param = 0.1; double method_param = 0.1;
int max_iter = 1000; int max_iter = 1000;
#endif #endif
mlp.train( &train_data, new_responses, 0, 0, mlp.train( &train_data, new_responses, 0, 0,
CvANN_MLP_TrainParams(cvTermCriteria(CV_TERMCRIT_ITER,max_iter,0.01), ANN_MLP::Params(TC(max_iter,0), method, method_param));
method, method_param));
model = ANN_MLP::create() mlp.create( &layer_sizes );
printf( "Training the classifier (may take a few minutes)...\n");
cvReleaseMat( &new_responses ); cvReleaseMat( &new_responses );
printf("\n"); printf("\n");
} }
mlp_response = cvCreateMat( 1, class_count, CV_32F ); Mat mlp_response;
// compute prediction error on train and test data // compute prediction error on train and test data
for( i = 0; i < nsamples_all; i++ ) for( i = 0; i < nsamples_all; i++ )
...@@ -481,38 +422,26 @@ int build_mlp_classifier( char* data_filename, ...@@ -481,38 +422,26 @@ int build_mlp_classifier( char* data_filename,
printf( "Recognition rate: train = %.1f%%, test = %.1f%%\n", printf( "Recognition rate: train = %.1f%%, test = %.1f%%\n",
train_hr*100., test_hr*100. ); train_hr*100., test_hr*100. );
// Save classifier to file if needed if( !filename_to_save.empty() )
if( filename_to_save ) model->save( filename_to_save );
mlp.save( filename_to_save );
cvReleaseMat( &mlp_response ); return true;
cvReleaseMat( &data );
cvReleaseMat( &responses );
return 0;
} }
static static bool
int build_knearest_classifier( char* data_filename, int K ) build_knearest_classifier( const string& data_filename, int K )
{ {
const int var_count = 16; const int var_count = 16;
CvMat* data = 0; Mat data;
CvMat train_data; CvMat train_data;
CvMat* responses; Mat responses;
int ok = read_num_class_data( data_filename, 16, &data, &responses );
int nsamples_all = 0, ntrain_samples = 0;
//int i, j;
//double /*train_hr = 0,*/ test_hr = 0;
CvANN_MLP mlp;
bool ok = read_num_class_data( data_filename, 16, &data, &responses );
if( !ok ) if( !ok )
{ return ok;
printf( "Could not read the database %s\n", data_filename );
return -1; int nsamples_all = 0, ntrain_samples = 0;
}
printf( "The database %s is loaded.\n", data_filename );
nsamples_all = data->rows; nsamples_all = data->rows;
ntrain_samples = (int)(nsamples_all*0.8); ntrain_samples = (int)(nsamples_all*0.8);
...@@ -521,12 +450,13 @@ int build_knearest_classifier( char* data_filename, int K ) ...@@ -521,12 +450,13 @@ int build_knearest_classifier( char* data_filename, int K )
cvGetRows( data, &train_data, 0, ntrain_samples ); cvGetRows( data, &train_data, 0, ntrain_samples );
// 2. train classifier // 2. train classifier
CvMat* train_resp = cvCreateMat( ntrain_samples, 1, CV_32FC1); Mat train_resp = cvCreateMat( ntrain_samples, 1, CV_32FC1);
for (int i = 0; i < ntrain_samples; i++) for (int i = 0; i < ntrain_samples; i++)
train_resp->data.fl[i] = responses->data.fl[i]; train_resp->data.fl[i] = responses->data.fl[i];
CvKNearest knearest(&train_data, train_resp); Ptr<KNearest> model = KNearest::create(true);
model->train(train_data, train_resp);
CvMat* nearests = cvCreateMat( (nsamples_all - ntrain_samples), K, CV_32FC1); Mat nearests = cvCreateMat( (nsamples_all - ntrain_samples), K, CV_32FC1);
float* _sample = new float[var_count * (nsamples_all - ntrain_samples)]; float* _sample = new float[var_count * (nsamples_all - ntrain_samples)];
CvMat sample = cvMat( nsamples_all - ntrain_samples, 16, CV_32FC1, _sample ); CvMat sample = cvMat( nsamples_all - ntrain_samples, 16, CV_32FC1, _sample );
float* true_results = new float[nsamples_all - ntrain_samples]; float* true_results = new float[nsamples_all - ntrain_samples];
...@@ -569,27 +499,20 @@ int build_knearest_classifier( char* data_filename, int K ) ...@@ -569,27 +499,20 @@ int build_knearest_classifier( char* data_filename, int K )
return 0; return 0;
} }
static static bool
int build_nbayes_classifier( char* data_filename ) build_nbayes_classifier( const string& data_filename )
{ {
const int var_count = 16; const int var_count = 16;
CvMat* data = 0; Mat data;
CvMat train_data; CvMat train_data;
CvMat* responses; Mat responses;
int ok = read_num_class_data( data_filename, 16, &data, &responses );
int nsamples_all = 0, ntrain_samples = 0;
//int i, j;
//double /*train_hr = 0, */test_hr = 0;
CvANN_MLP mlp;
bool ok = read_num_class_data( data_filename, 16, &data, &responses );
if( !ok ) if( !ok )
{ return ok;
printf( "Could not read the database %s\n", data_filename );
return -1; int nsamples_all = 0, ntrain_samples = 0;
}
printf( "The database %s is loaded.\n", data_filename );
nsamples_all = data->rows; nsamples_all = data->rows;
ntrain_samples = (int)(nsamples_all*0.5); ntrain_samples = (int)(nsamples_all*0.5);
...@@ -598,7 +521,7 @@ int build_nbayes_classifier( char* data_filename ) ...@@ -598,7 +521,7 @@ int build_nbayes_classifier( char* data_filename )
cvGetRows( data, &train_data, 0, ntrain_samples ); cvGetRows( data, &train_data, 0, ntrain_samples );
// 2. train classifier // 2. train classifier
CvMat* train_resp = cvCreateMat( ntrain_samples, 1, CV_32FC1); Mat train_resp = cvCreateMat( ntrain_samples, 1, CV_32FC1);
for (int i = 0; i < ntrain_samples; i++) for (int i = 0; i < ntrain_samples; i++)
train_resp->data.fl[i] = responses->data.fl[i]; train_resp->data.fl[i] = responses->data.fl[i];
CvNormalBayesClassifier nbayes(&train_data, train_resp); CvNormalBayesClassifier nbayes(&train_data, train_resp);
...@@ -638,23 +561,23 @@ int build_nbayes_classifier( char* data_filename ) ...@@ -638,23 +561,23 @@ int build_nbayes_classifier( char* data_filename )
return 0; return 0;
} }
static static bool
int build_svm_classifier( char* data_filename, const char* filename_to_save, const char* filename_to_load ) build_svm_classifier( const string& data_filename,
const string& filename_to_save,
const string& filename_to_load )
{ {
CvMat* data = 0; Mat data;
CvMat* responses = 0; Mat responses;
CvMat* train_resp = 0; Mat train_resp;
CvMat train_data; CvMat train_data;
int nsamples_all = 0, ntrain_samples = 0; int nsamples_all = 0, ntrain_samples = 0;
int var_count; int var_count;
CvSVM svm; Ptr<SVM> model;
int ok = read_num_class_data( data_filename, 16, &data, &responses ); bool ok = read_num_class_data( data_filename, 16, &data, &responses );
if( !ok ) if( !ok )
{ return ok;
printf( "Could not read the database %s\n", data_filename );
return -1;
}
////////// SVM parameters /////////////////////////////// ////////// SVM parameters ///////////////////////////////
CvSVMParams param; CvSVMParams param;
param.kernel_type=CvSVM::LINEAR; param.kernel_type=CvSVM::LINEAR;
...@@ -722,15 +645,10 @@ int build_svm_classifier( char* data_filename, const char* filename_to_save, con ...@@ -722,15 +645,10 @@ int build_svm_classifier( char* data_filename, const char* filename_to_save, con
printf("true_resp = %f%%\n", (float)true_resp / (nsamples_all - ntrain_samples) * 100); printf("true_resp = %f%%\n", (float)true_resp / (nsamples_all - ntrain_samples) * 100);
if( filename_to_save ) if( !filename_to_save.empty() )
svm.save( filename_to_save ); model->save( filename_to_save );
cvReleaseMat( &train_resp ); return true;
cvReleaseMat( &result );
cvReleaseMat( &data );
cvReleaseMat( &responses );
return 0;
} }
int main( int argc, char *argv[] ) int main( int argc, char *argv[] )
...@@ -768,17 +686,17 @@ int main( int argc, char *argv[] ) ...@@ -768,17 +686,17 @@ int main( int argc, char *argv[] )
method = 2; method = 2;
} }
else if ( strcmp(argv[i], "-knearest") == 0) else if ( strcmp(argv[i], "-knearest") == 0)
{ {
method = 3; method = 3;
} }
else if ( strcmp(argv[i], "-nbayes") == 0) else if ( strcmp(argv[i], "-nbayes") == 0)
{ {
method = 4; method = 4;
} }
else if ( strcmp(argv[i], "-svm") == 0) else if ( strcmp(argv[i], "-svm") == 0)
{ {
method = 5; method = 5;
} }
else else
break; break;
} }
......
samples/cpp/points_classifier.cpp
View file @ 10b60f8d
...@@ -229,22 +229,7 @@ static void find_decision_boundary_ANN( const Mat& layer_sizes ) ...@@ -229,22 +229,7 @@ static void find_decision_boundary_ANN( const Mat& layer_sizes )
Ptr<TrainData> tdata = TrainData::create(samples, ROW_SAMPLE, trainClasses); Ptr<TrainData> tdata = TrainData::create(samples, ROW_SAMPLE, trainClasses);
ann->train(tdata); ann->train(tdata);
predict_and_paint(ann, imgDst);
Mat testSample( 1, 2, CV_32FC1 );
Mat outputs;
for( int y = 0; y < img.rows; y += testStep )
{
for( int x = 0; x < img.cols; x += testStep )
{
testSample.at<float>(0) = (float)x;
testSample.at<float>(1) = (float)y;
ann->predict( testSample, outputs );
Point maxLoc;
minMaxLoc( outputs, 0, 0, 0, &maxLoc );
imgDst.at<Vec3b>(y, x) = classColors[maxLoc.x];
}
}
} }
#endif #endif
......
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