opencv/samples/cpp/letter_recog.cpp

559 lines
17 KiB
C++
Raw Normal View History

2016-02-15 21:37:29 +08:00
#include "opencv2/core.hpp"
#include "opencv2/ml.hpp"
2010-11-28 07:15:16 +08:00
2010-11-29 17:31:47 +08:00
#include <cstdio>
2011-04-29 20:41:30 +08:00
#include <vector>
#include <iostream>
2010-11-28 07:15:16 +08:00
using namespace std;
using namespace cv;
using namespace cv::ml;
2010-11-28 07:15:16 +08:00
static void help()
2010-12-04 16:30:46 +08:00
{
printf("\nThe sample demonstrates how to train Random Trees classifier\n"
"(or Boosting classifier, or MLP, or Knearest, or Nbayes, or Support Vector Machines - see main()) using the provided dataset.\n"
"\n"
"We use the sample database letter-recognition.data\n"
"from UCI Repository, here is the link:\n"
"\n"
"Newman, D.J. & Hettich, S. & Blake, C.L. & Merz, C.J. (1998).\n"
"UCI Repository of machine learning databases\n"
"[http://www.ics.uci.edu/~mlearn/MLRepository.html].\n"
"Irvine, CA: University of California, Department of Information and Computer Science.\n"
"\n"
"The dataset consists of 20000 feature vectors along with the\n"
"responses - capital latin letters A..Z.\n"
"The first 16000 (10000 for boosting)) samples are used for training\n"
"and the remaining 4000 (10000 for boosting) - to test the classifier.\n"
"======================================================\n");
2010-12-04 16:30:46 +08:00
printf("\nThis is letter recognition sample.\n"
2015-08-01 23:24:23 +08:00
"The usage: letter_recog [-data=<path to letter-recognition.data>] \\\n"
" [-save=<output XML file for the classifier>] \\\n"
" [-load=<XML file with the pre-trained classifier>] \\\n"
" [-boost|-mlp|-knearest|-nbayes|-svm] # to use boost/mlp/knearest/SVM classifier instead of default Random Trees\n" );
2010-12-04 16:30:46 +08:00
}
2010-11-28 07:15:16 +08:00
// This function reads data and responses from the file <filename>
static bool
read_num_class_data( const string& filename, int var_count,
Mat* _data, Mat* _responses )
2010-11-28 07:15:16 +08:00
{
const int M = 1024;
char buf[M+2];
Mat el_ptr(1, var_count, CV_32F);
int i;
vector<int> responses;
2010-11-28 07:15:16 +08:00
_data->release();
_responses->release();
FILE* f = fopen( filename.c_str(), "rt" );
if( !f )
{
cout << "Could not read the database " << filename << endl;
return false;
}
2010-11-28 07:15:16 +08:00
for(;;)
{
char* ptr;
if( !fgets( buf, M, f ) || !strchr( buf, ',' ) )
break;
responses.push_back((int)buf[0]);
2010-11-28 07:15:16 +08:00
ptr = buf+2;
for( i = 0; i < var_count; i++ )
2010-11-28 07:15:16 +08:00
{
int n = 0;
sscanf( ptr, "%f%n", &el_ptr.at<float>(i), &n );
2010-11-28 07:15:16 +08:00
ptr += n + 1;
}
if( i < var_count )
2010-11-28 07:15:16 +08:00
break;
_data->push_back(el_ptr);
2010-11-28 07:15:16 +08:00
}
fclose(f);
Mat(responses).copyTo(*_responses);
2010-11-28 07:15:16 +08:00
cout << "The database " << filename << " is loaded.\n";
2010-11-28 07:15:16 +08:00
return true;
2010-11-28 07:15:16 +08:00
}
template<typename T>
static Ptr<T> load_classifier(const string& filename_to_load)
2010-11-28 07:15:16 +08:00
{
// load classifier from the specified file
Ptr<T> model = StatModel::load<T>( filename_to_load );
if( model.empty() )
cout << "Could not read the classifier " << filename_to_load << endl;
else
cout << "The classifier " << filename_to_load << " is loaded.\n";
2010-11-28 07:15:16 +08:00
return model;
}
2010-11-28 07:15:16 +08:00
static Ptr<TrainData>
prepare_train_data(const Mat& data, const Mat& responses, int ntrain_samples)
{
Mat sample_idx = Mat::zeros( 1, data.rows, CV_8U );
Mat train_samples = sample_idx.colRange(0, ntrain_samples);
train_samples.setTo(Scalar::all(1));
2010-11-28 07:15:16 +08:00
int nvars = data.cols;
Mat var_type( nvars + 1, 1, CV_8U );
var_type.setTo(Scalar::all(VAR_ORDERED));
var_type.at<uchar>(nvars) = VAR_CATEGORICAL;
2010-11-28 07:15:16 +08:00
return TrainData::create(data, ROW_SAMPLE, responses,
noArray(), sample_idx, noArray(), var_type);
}
2010-11-28 07:15:16 +08:00
inline TermCriteria TC(int iters, double eps)
{
return TermCriteria(TermCriteria::MAX_ITER + (eps > 0 ? TermCriteria::EPS : 0), iters, eps);
}
2010-11-28 07:15:16 +08:00
static void test_and_save_classifier(const Ptr<StatModel>& model,
const Mat& data, const Mat& responses,
int ntrain_samples, int rdelta,
const string& filename_to_save)
{
int i, nsamples_all = data.rows;
double train_hr = 0, test_hr = 0;
2010-11-28 07:15:16 +08:00
// compute prediction error on train and test data
for( i = 0; i < nsamples_all; i++ )
{
Mat sample = data.row(i);
2010-11-28 07:15:16 +08:00
float r = model->predict( sample );
2014-08-04 18:22:31 +08:00
r = std::abs(r + rdelta - responses.at<int>(i)) <= FLT_EPSILON ? 1.f : 0.f;
2010-11-28 07:15:16 +08:00
if( i < ntrain_samples )
train_hr += r;
else
test_hr += r;
}
test_hr /= nsamples_all - ntrain_samples;
train_hr = ntrain_samples > 0 ? train_hr/ntrain_samples : 1.;
2010-11-28 07:15:16 +08:00
printf( "Recognition rate: train = %.1f%%, test = %.1f%%\n",
train_hr*100., test_hr*100. );
if( !filename_to_save.empty() )
2010-11-28 07:15:16 +08:00
{
model->save( filename_to_save );
2010-11-28 07:15:16 +08:00
}
}
2010-11-28 07:15:16 +08:00
static bool
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() )
2010-11-28 07:15:16 +08:00
{
model = load_classifier<RTrees>(filename_to_load);
if( model.empty() )
return false;
ntrain_samples = 0;
}
else
{
// create classifier by using <data> and <responses>
cout << "Training the classifier ...\n";
// Params( int maxDepth, int minSampleCount,
// double regressionAccuracy, bool useSurrogates,
// int maxCategories, const Mat& priors,
// bool calcVarImportance, int nactiveVars,
// TermCriteria termCrit );
Ptr<TrainData> tdata = prepare_train_data(data, responses, ntrain_samples);
model = RTrees::create();
model->setMaxDepth(10);
model->setMinSampleCount(10);
model->setRegressionAccuracy(0);
model->setUseSurrogates(false);
model->setMaxCategories(15);
model->setPriors(Mat());
model->setCalculateVarImportance(true);
model->setActiveVarCount(4);
model->setTermCriteria(TC(100,0.01f));
model->train(tdata);
cout << endl;
2010-11-28 07:15:16 +08:00
}
test_and_save_classifier(model, data, responses, ntrain_samples, 0, filename_to_save);
cout << "Number of trees: " << model->getRoots().size() << endl;
2010-11-28 07:15:16 +08:00
// Print variable importance
Mat var_importance = model->getVarImportance();
if( !var_importance.empty() )
{
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);
}
2010-11-28 07:15:16 +08:00
return true;
2010-11-28 07:15:16 +08:00
}
static bool
build_boost_classifier( const string& data_filename,
const string& filename_to_save,
const string& filename_to_load )
2010-11-28 07:15:16 +08:00
{
const int class_count = 26;
Mat data;
Mat responses;
Mat weak_responses;
2010-11-28 07:15:16 +08:00
bool ok = read_num_class_data( data_filename, 16, &data, &responses );
2010-11-28 07:15:16 +08:00
if( !ok )
return ok;
2010-11-28 07:15:16 +08:00
int i, j, k;
Ptr<Boost> model;
int nsamples_all = data.rows;
int ntrain_samples = (int)(nsamples_all*0.5);
int var_count = data.cols;
2010-11-28 07:15:16 +08:00
// Create or load Boosted Tree classifier
if( !filename_to_load.empty() )
2010-11-28 07:15:16 +08:00
{
model = load_classifier<Boost>(filename_to_load);
if( model.empty() )
return false;
2010-11-28 07:15:16 +08:00
ntrain_samples = 0;
}
else
{
// !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
//
// As currently boosted tree classifier in MLL can only be trained
// for 2-class problems, we transform the training database by
// "unrolling" each training sample as many times as the number of
// classes (26) that we have.
//
// !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
Mat new_data( ntrain_samples*class_count, var_count + 1, CV_32F );
Mat new_responses( ntrain_samples*class_count, 1, CV_32S );
2010-11-28 07:15:16 +08:00
// 1. unroll the database type mask
printf( "Unrolling the database...\n");
for( i = 0; i < ntrain_samples; i++ )
{
const float* data_row = data.ptr<float>(i);
2010-11-28 07:15:16 +08:00
for( j = 0; j < class_count; j++ )
{
float* new_data_row = (float*)new_data.ptr<float>(i*class_count+j);
memcpy(new_data_row, data_row, var_count*sizeof(data_row[0]));
2010-11-28 07:15:16 +08:00
new_data_row[var_count] = (float)j;
new_responses.at<int>(i*class_count + j) = responses.at<int>(i) == j+'A';
2010-11-28 07:15:16 +08:00
}
}
Mat var_type( 1, var_count + 2, CV_8U );
var_type.setTo(Scalar::all(VAR_ORDERED));
var_type.at<uchar>(var_count) = var_type.at<uchar>(var_count+1) = VAR_CATEGORICAL;
2010-11-28 07:15:16 +08:00
Ptr<TrainData> tdata = TrainData::create(new_data, ROW_SAMPLE, new_responses,
noArray(), noArray(), noArray(), var_type);
vector<double> priors(2);
priors[0] = 1;
priors[1] = 26;
cout << "Training the classifier (may take a few minutes)...\n";
model = Boost::create();
model->setBoostType(Boost::GENTLE);
model->setWeakCount(100);
model->setWeightTrimRate(0.95);
model->setMaxDepth(5);
model->setUseSurrogates(false);
model->setPriors(Mat(priors));
model->train(tdata);
cout << endl;
2010-11-28 07:15:16 +08:00
}
Mat temp_sample( 1, var_count + 1, CV_32F );
float* tptr = temp_sample.ptr<float>();
2010-11-28 07:15:16 +08:00
// compute prediction error on train and test data
double train_hr = 0, test_hr = 0;
2010-11-28 07:15:16 +08:00
for( i = 0; i < nsamples_all; i++ )
{
int best_class = 0;
double max_sum = -DBL_MAX;
const float* ptr = data.ptr<float>(i);
2010-11-28 07:15:16 +08:00
for( k = 0; k < var_count; k++ )
tptr[k] = ptr[k];
2010-11-28 07:15:16 +08:00
for( j = 0; j < class_count; j++ )
{
tptr[var_count] = (float)j;
float s = model->predict( temp_sample, noArray(), StatModel::RAW_OUTPUT );
if( max_sum < s )
2010-11-28 07:15:16 +08:00
{
max_sum = s;
2010-11-28 07:15:16 +08:00
best_class = j + 'A';
}
}
double r = std::abs(best_class - responses.at<int>(i)) < FLT_EPSILON ? 1 : 0;
2010-11-28 07:15:16 +08:00
if( i < ntrain_samples )
train_hr += r;
else
test_hr += r;
}
test_hr /= nsamples_all-ntrain_samples;
train_hr = ntrain_samples > 0 ? train_hr/ntrain_samples : 1.;
2010-11-28 07:15:16 +08:00
printf( "Recognition rate: train = %.1f%%, test = %.1f%%\n",
train_hr*100., test_hr*100. );
cout << "Number of trees: " << model->getRoots().size() << endl;
2010-11-28 07:15:16 +08:00
// Save classifier to file if needed
if( !filename_to_save.empty() )
model->save( filename_to_save );
2010-11-28 07:15:16 +08:00
return true;
2010-11-28 07:15:16 +08:00
}
static bool
build_mlp_classifier( const string& data_filename,
const string& filename_to_save,
const string& filename_to_load )
2010-11-28 07:15:16 +08:00
{
const int class_count = 26;
Mat data;
Mat responses;
2010-11-28 07:15:16 +08:00
bool ok = read_num_class_data( data_filename, 16, &data, &responses );
2010-11-28 07:15:16 +08:00
if( !ok )
return ok;
2010-11-28 07:15:16 +08:00
Ptr<ANN_MLP> model;
int nsamples_all = data.rows;
int ntrain_samples = (int)(nsamples_all*0.8);
2010-11-28 07:15:16 +08:00
// Create or load MLP classifier
if( !filename_to_load.empty() )
2010-11-28 07:15:16 +08:00
{
model = load_classifier<ANN_MLP>(filename_to_load);
if( model.empty() )
return false;
2010-11-28 07:15:16 +08:00
ntrain_samples = 0;
}
else
{
// !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
//
// MLP does not support categorical variables by explicitly.
// So, instead of the output class label, we will use
// a binary vector of <class_count> components for training and,
// therefore, MLP will give us a vector of "probabilities" at the
// prediction stage
//
// !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
Mat train_data = data.rowRange(0, ntrain_samples);
Mat train_responses = Mat::zeros( ntrain_samples, class_count, CV_32F );
2010-11-28 07:15:16 +08:00
// 1. unroll the responses
cout << "Unrolling the responses...\n";
for( int i = 0; i < ntrain_samples; i++ )
2010-11-28 07:15:16 +08:00
{
int cls_label = responses.at<int>(i) - 'A';
train_responses.at<float>(i, cls_label) = 1.f;
2010-11-28 07:15:16 +08:00
}
// 2. train classifier
int layer_sz[] = { data.cols, 100, 100, class_count };
int nlayers = (int)(sizeof(layer_sz)/sizeof(layer_sz[0]));
Mat layer_sizes( 1, nlayers, CV_32S, layer_sz );
2012-10-17 07:18:30 +08:00
2010-11-28 07:15:16 +08:00
#if 1
int method = ANN_MLP::BACKPROP;
double method_param = 0.001;
int max_iter = 300;
2010-11-28 07:15:16 +08:00
#else
int method = ANN_MLP::RPROP;
double method_param = 0.1;
int max_iter = 1000;
2010-11-28 07:15:16 +08:00
#endif
2012-10-17 07:18:30 +08:00
Ptr<TrainData> tdata = TrainData::create(train_data, ROW_SAMPLE, train_responses);
2010-11-28 07:15:16 +08:00
cout << "Training the classifier (may take a few minutes)...\n";
model = ANN_MLP::create();
model->setLayerSizes(layer_sizes);
model->setActivationFunction(ANN_MLP::SIGMOID_SYM, 0, 0);
model->setTermCriteria(TC(max_iter,0));
model->setTrainMethod(method, method_param);
model->train(tdata);
cout << endl;
2010-11-28 07:15:16 +08:00
}
test_and_save_classifier(model, data, responses, ntrain_samples, 'A', filename_to_save);
return true;
2010-11-28 07:15:16 +08:00
}
static bool
build_knearest_classifier( const string& data_filename, int K )
{
Mat data;
Mat responses;
bool ok = read_num_class_data( data_filename, 16, &data, &responses );
if( !ok )
return ok;
int nsamples_all = data.rows;
int ntrain_samples = (int)(nsamples_all*0.8);
2011-04-16 03:46:45 +08:00
// create classifier by using <data> and <responses>
cout << "Training the classifier ...\n";
Ptr<TrainData> tdata = prepare_train_data(data, responses, ntrain_samples);
Ptr<KNearest> model = KNearest::create();
model->setDefaultK(K);
model->setIsClassifier(true);
model->train(tdata);
cout << endl;
test_and_save_classifier(model, data, responses, ntrain_samples, 0, string());
return true;
}
static bool
build_nbayes_classifier( const string& data_filename )
{
Mat data;
Mat responses;
bool ok = read_num_class_data( data_filename, 16, &data, &responses );
if( !ok )
return ok;
Ptr<NormalBayesClassifier> model;
int nsamples_all = data.rows;
int ntrain_samples = (int)(nsamples_all*0.8);
2011-04-16 03:46:45 +08:00
// create classifier by using <data> and <responses>
cout << "Training the classifier ...\n";
Ptr<TrainData> tdata = prepare_train_data(data, responses, ntrain_samples);
model = NormalBayesClassifier::create();
model->train(tdata);
cout << endl;
test_and_save_classifier(model, data, responses, ntrain_samples, 0, string());
return true;
}
2010-11-28 07:15:16 +08:00
static bool
build_svm_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<SVM> 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() )
{
model = load_classifier<SVM>(filename_to_load);
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);
model = SVM::create();
model->setType(SVM::C_SVC);
model->setKernel(SVM::LINEAR);
model->setC(1);
model->train(tdata);
cout << endl;
}
test_and_save_classifier(model, data, responses, ntrain_samples, 0, filename_to_save);
return true;
}
2010-11-28 07:15:16 +08:00
int main( int argc, char *argv[] )
{
string filename_to_save = "";
string filename_to_load = "";
2015-08-01 23:24:23 +08:00
string data_filename;
2010-11-28 07:15:16 +08:00
int method = 0;
2018-10-31 20:48:56 +08:00
cv::CommandLineParser parser(argc, argv, "{data|letter-recognition.data|}{save||}{load||}{boost||}"
2017-07-26 13:39:53 +08:00
"{mlp||}{knn knearest||}{nbayes||}{svm||}");
2018-10-31 20:48:56 +08:00
data_filename = samples::findFile(parser.get<string>("data"));
2015-08-01 23:24:23 +08:00
if (parser.has("save"))
filename_to_save = parser.get<string>("save");
if (parser.has("load"))
2018-10-31 20:48:56 +08:00
filename_to_load = samples::findFile(parser.get<string>("load"));
2015-08-01 23:24:23 +08:00
if (parser.has("boost"))
method = 1;
else if (parser.has("mlp"))
method = 2;
else if (parser.has("knearest"))
method = 3;
else if (parser.has("nbayes"))
method = 4;
else if (parser.has("svm"))
method = 5;
2017-07-26 13:39:53 +08:00
help();
2015-08-01 23:24:23 +08:00
if( (method == 0 ?
2010-11-28 07:15:16 +08:00
build_rtrees_classifier( data_filename, filename_to_save, filename_to_load ) :
method == 1 ?
build_boost_classifier( data_filename, filename_to_save, filename_to_load ) :
method == 2 ?
build_mlp_classifier( data_filename, filename_to_save, filename_to_load ) :
method == 3 ?
build_knearest_classifier( data_filename, 10 ) :
method == 4 ?
build_nbayes_classifier( data_filename) :
method == 5 ?
build_svm_classifier( data_filename, filename_to_save, filename_to_load ):
2010-11-28 07:15:16 +08:00
-1) < 0)
2017-07-26 13:39:53 +08:00
2010-11-28 07:15:16 +08:00
return 0;
}