#include <trainingsampleset.h>

Public Member Functions
	TrainingSampleSet (const FontInfoTable &fontinfo_table)

	~TrainingSampleSet ()

bool	Serialize (FILE *fp) const

bool	DeSerialize (bool swap, FILE *fp)

int	num_samples () const

int	num_raw_samples () const

int	NumFonts () const

const UNICHARSET &	unicharset () const

int	charsetsize () const

const FontInfoTable &	fontinfo_table () const

void	LoadUnicharset (const char *filename)

int	AddSample (const char unichar, TrainingSample sample)

void	AddSample (int unichar_id, TrainingSample *sample)

int	NumClassSamples (int font_id, int class_id, bool randomize) const

const TrainingSample *	GetSample (int index) const

const TrainingSample *	GetSample (int font_id, int class_id, int index) const

TrainingSample *	MutableSample (int font_id, int class_id, int index)

STRING	SampleToString (const TrainingSample &sample) const

const BitVector &	GetCloudFeatures (int font_id, int class_id) const

const GenericVector< int > &	GetCanonicalFeatures (int font_id, int class_id) const

float	UnicharDistance (const UnicharAndFonts &uf1, const UnicharAndFonts &uf2, bool matched_fonts, const IntFeatureMap &feature_map)

float	ClusterDistance (int font_id1, int class_id1, int font_id2, int class_id2, const IntFeatureMap &feature_map)

float	ComputeClusterDistance (int font_id1, int class_id1, int font_id2, int class_id2, const IntFeatureMap &feature_map) const

int	ReliablySeparable (int font_id1, int class_id1, int font_id2, int class_id2, const IntFeatureMap &feature_map, bool thorough) const

int	GlobalSampleIndex (int font_id, int class_id, int index) const

const TrainingSample *	GetCanonicalSample (int font_id, int class_id) const

float	GetCanonicalDist (int font_id, int class_id) const

TrainingSample *	mutable_sample (int index)

TrainingSample *	extract_sample (int index)

void	IndexFeatures (const IntFeatureSpace &feature_space)

void	KillSample (TrainingSample *sample)

void	DeleteDeadSamples ()

bool	DeleteableSample (const TrainingSample *sample)

void	OrganizeByFontAndClass ()

void	SetupFontIdMap ()

void	ComputeCanonicalSamples (const IntFeatureMap &map, bool debug)

void	ReplicateAndRandomizeSamples ()

void	ComputeCanonicalFeatures ()

void	ComputeCloudFeatures (int feature_space_size)

void	AddAllFontsForClass (int class_id, Shape *shape) const

void	DisplaySamplesWithFeature (int f_index, const Shape &shape, const IntFeatureSpace &feature_space, ScrollView::Color color, ScrollView *window) const

Detailed Description

Definition at line 43 of file trainingsampleset.h.

Constructor & Destructor Documentation

◆ TrainingSampleSet()

tesseract::TrainingSampleSet::TrainingSampleSet ( const FontInfoTable & fontinfo_table )

explicit

Definition at line 70 of file trainingsampleset.cpp.

   : num_raw_samples_(0), unicharset_size_(0),
     font_class_array_(NULL), fontinfo_table_(font_table) {
 }

◆ ~TrainingSampleSet()

tesseract::TrainingSampleSet::~TrainingSampleSet ( )

Definition at line 75 of file trainingsampleset.cpp.

                                       {
   delete font_class_array_;
 }

Member Function Documentation

◆ AddAllFontsForClass()

void tesseract::TrainingSampleSet::AddAllFontsForClass	(	int	class_id,
		Shape *	shape
	)		const

Definition at line 734 of file trainingsampleset.cpp.

                                                                             {
   for (int f = 0; f < font_id_map_.CompactSize(); ++f) {
     int font_id = font_id_map_.CompactToSparse(f);
     shape->AddToShape(class_id, font_id);
   }
 }

◆ AddSample() [1/2]

int tesseract::TrainingSampleSet::AddSample	(	const char *	unichar,
		TrainingSample *	sample
	)

Definition at line 129 of file trainingsampleset.cpp.

                                                                             {
   if (!unicharset_.contains_unichar(unichar)) {
     unicharset_.unichar_insert(unichar);
     if (unicharset_.size() > MAX_NUM_CLASSES) {
       tprintf("Error: Size of unicharset in TrainingSampleSet::AddSample is "
               "greater than MAX_NUM_CLASSES\n");
       return -1;
     }
   }
   UNICHAR_ID char_id = unicharset_.unichar_to_id(unichar);
   AddSample(char_id, sample);
   return char_id;
 }

◆ AddSample() [2/2]

void tesseract::TrainingSampleSet::AddSample	(	int	unichar_id,
		TrainingSample *	sample
	)

Definition at line 145 of file trainingsampleset.cpp.

                                                                         {
   sample->set_class_id(unichar_id);
   samples_.push_back(sample);
   num_raw_samples_ = samples_.size();
   unicharset_size_ = unicharset_.size();
 }

◆ charsetsize()

int tesseract::TrainingSampleSet::charsetsize ( ) const

inline

Definition at line 67 of file trainingsampleset.h.

                           {
     return unicharset_size_;
   }

◆ ClusterDistance()

float tesseract::TrainingSampleSet::ClusterDistance	(	int	font_id1,
		int	class_id1,
		int	font_id2,
		int	class_id2,
		const IntFeatureMap &	feature_map
	)

Definition at line 296 of file trainingsampleset.cpp.

                                                                            {
   ASSERT_HOST(font_class_array_ != NULL);
   int font_index1 = font_id_map_.SparseToCompact(font_id1);
   int font_index2 = font_id_map_.SparseToCompact(font_id2);
   if (font_index1 < 0 || font_index2 < 0)
     return 0.0f;
   FontClassInfo& fc_info = (*font_class_array_)(font_index1, class_id1);
   if (font_id1 == font_id2) {
     // Special case cache for speed.
     if (fc_info.unichar_distance_cache.size() == 0)
       fc_info.unichar_distance_cache.init_to_size(unicharset_size_, -1.0f);
     if (fc_info.unichar_distance_cache[class_id2] < 0) {
       // Distance has to be calculated.
       float result = ComputeClusterDistance(font_id1, class_id1,
                                             font_id2, class_id2,
                                             feature_map);
       fc_info.unichar_distance_cache[class_id2] = result;
       // Copy to the symmetric cache entry.
       FontClassInfo& fc_info2 = (*font_class_array_)(font_index2, class_id2);
       if (fc_info2.unichar_distance_cache.size() == 0)
         fc_info2.unichar_distance_cache.init_to_size(unicharset_size_, -1.0f);
       fc_info2.unichar_distance_cache[class_id1] = result;
     }
     return fc_info.unichar_distance_cache[class_id2];
   } else if (class_id1 == class_id2) {
     // Another special-case cache for equal class-id.
     if (fc_info.font_distance_cache.size() == 0)
       fc_info.font_distance_cache.init_to_size(font_id_map_.CompactSize(),
                                                -1.0f);
     if (fc_info.font_distance_cache[font_index2] < 0) {
       // Distance has to be calculated.
       float result = ComputeClusterDistance(font_id1, class_id1,
                                             font_id2, class_id2,
                                             feature_map);
       fc_info.font_distance_cache[font_index2] = result;
       // Copy to the symmetric cache entry.
       FontClassInfo& fc_info2 = (*font_class_array_)(font_index2, class_id2);
       if (fc_info2.font_distance_cache.size() == 0)
         fc_info2.font_distance_cache.init_to_size(font_id_map_.CompactSize(),
                                                   -1.0f);
       fc_info2.font_distance_cache[font_index1] = result;
     }
     return fc_info.font_distance_cache[font_index2];
   }
   // Both font and class are different. Linear search for class_id2/font_id2
   // in what is a hopefully short list of distances.
   int cache_index = 0;
   while (cache_index < fc_info.distance_cache.size() &&
          (fc_info.distance_cache[cache_index].unichar_id != class_id2 ||
           fc_info.distance_cache[cache_index].font_id != font_id2))
     ++cache_index;
   if (cache_index == fc_info.distance_cache.size()) {
     // Distance has to be calculated.
     float result = ComputeClusterDistance(font_id1, class_id1,
                                           font_id2, class_id2,
                                           feature_map);
     FontClassDistance fc_dist = { class_id2, font_id2, result };
     fc_info.distance_cache.push_back(fc_dist);
     // Copy to the symmetric cache entry. We know it isn't there already, as
     // we always copy to the symmetric entry.
     FontClassInfo& fc_info2 = (*font_class_array_)(font_index2, class_id2);
     fc_dist.unichar_id = class_id1;
     fc_dist.font_id = font_id1;
     fc_info2.distance_cache.push_back(fc_dist);
   }
   return fc_info.distance_cache[cache_index].distance;
 }

◆ ComputeCanonicalFeatures()

void tesseract::TrainingSampleSet::ComputeCanonicalFeatures ( )

Definition at line 694 of file trainingsampleset.cpp.

                                                  {
   ASSERT_HOST(font_class_array_ != NULL);
   int font_size = font_id_map_.CompactSize();
   for (int font_index = 0; font_index < font_size; ++font_index) {
     int font_id = font_id_map_.CompactToSparse(font_index);
     for (int c = 0; c < unicharset_size_; ++c) {
       int num_samples = NumClassSamples(font_id, c, false);
       if (num_samples == 0)
         continue;
       const TrainingSample* sample = GetCanonicalSample(font_id, c);
       FontClassInfo& fcinfo = (*font_class_array_)(font_index, c);
       fcinfo.canonical_features = sample->indexed_features();
     }
   }
 }

◆ ComputeCanonicalSamples()

void tesseract::TrainingSampleSet::ComputeCanonicalSamples	(	const IntFeatureMap &	map,
		bool	debug
	)

Definition at line 569 of file trainingsampleset.cpp.

                                                             {
   ASSERT_HOST(font_class_array_ != NULL);
   IntFeatureDist f_table;
   if (debug) tprintf("feature table size %d\n", map.sparse_size());
   f_table.Init(&map);
   int worst_s1 = 0;
   int worst_s2 = 0;
   double global_worst_dist = 0.0;
   // Compute distances independently for each font and char index.
   int font_size = font_id_map_.CompactSize();
   for (int font_index = 0; font_index < font_size; ++font_index) {
     int font_id = font_id_map_.CompactToSparse(font_index);
     for (int c = 0; c < unicharset_size_; ++c) {
       int samples_found = 0;
       FontClassInfo& fcinfo = (*font_class_array_)(font_index, c);
       if (fcinfo.samples.size() == 0 ||
           (kTestChar >= 0 && c != kTestChar)) {
         fcinfo.canonical_sample = -1;
         fcinfo.canonical_dist = 0.0f;
         if (debug) tprintf("Skipping class %d\n", c);
         continue;
       }
       // The canonical sample will be the one with the min_max_dist, which
       // is the sample with the lowest maximum distance to all other samples.
       double min_max_dist = 2.0;
       // We keep track of the farthest apart pair (max_s1, max_s2) which
       // are max_max_dist apart, so we can see how bad the variability is.
       double max_max_dist = 0.0;
       int max_s1 = 0;
       int max_s2 = 0;
       fcinfo.canonical_sample = fcinfo.samples[0];
       fcinfo.canonical_dist = 0.0f;
       for (int i = 0; i < fcinfo.samples.size(); ++i) {
         int s1 = fcinfo.samples[i];
         const GenericVector<int>& features1 = samples_[s1]->indexed_features();
         f_table.Set(features1, features1.size(), true);
         double max_dist = 0.0;
         // Run the full squared-order search for similar samples. It is still
         // reasonably fast because f_table.FeatureDistance is fast, but we
         // may have to reconsider if we start playing with too many samples
         // of a single char/font.
         for (int j = 0; j < fcinfo.samples.size(); ++j) {
           int s2 = fcinfo.samples[j];
           if (samples_[s2]->class_id() != c  ||
               samples_[s2]->font_id() != font_id ||
               s2 == s1)
             continue;
           GenericVector<int> features2 = samples_[s2]->indexed_features();
           double dist = f_table.FeatureDistance(features2);
           if (dist > max_dist) {
             max_dist = dist;
             if (dist > max_max_dist) {
               max_s1 = s1;
               max_s2 = s2;
             }
           }
         }
         // Using Set(..., false) is far faster than re initializing, due to
         // the sparseness of the feature space.
         f_table.Set(features1, features1.size(), false);
         samples_[s1]->set_max_dist(max_dist);
         ++samples_found;
         if (max_dist < min_max_dist) {
           fcinfo.canonical_sample = s1;
           fcinfo.canonical_dist = max_dist;
         }
         UpdateRange(max_dist, &min_max_dist, &max_max_dist);
       }
       if (max_max_dist > global_worst_dist) {
         // Keep a record of the worst pair over all characters/fonts too.
         global_worst_dist = max_max_dist;
         worst_s1 = max_s1;
         worst_s2 = max_s2;
       }
       if (debug) {
         tprintf("Found %d samples of class %d=%s, font %d, "
                 "dist range [%g, %g], worst pair= %s, %s\n",
                 samples_found, c, unicharset_.debug_str(c).string(),
                 font_index, min_max_dist, max_max_dist,
                 SampleToString(*samples_[max_s1]).string(),
                 SampleToString(*samples_[max_s2]).string());
       }
     }
   }
   if (debug) {
     tprintf("Global worst dist = %g, between sample %d and %d\n",
             global_worst_dist, worst_s1, worst_s2);
   }
 }

◆ ComputeCloudFeatures()

void tesseract::TrainingSampleSet::ComputeCloudFeatures ( int feature_space_size )

Definition at line 712 of file trainingsampleset.cpp.

                                                                    {
   ASSERT_HOST(font_class_array_ != NULL);
   int font_size = font_id_map_.CompactSize();
   for (int font_index = 0; font_index < font_size; ++font_index) {
     int font_id = font_id_map_.CompactToSparse(font_index);
     for (int c = 0; c < unicharset_size_; ++c) {
       int num_samples = NumClassSamples(font_id, c, false);
       if (num_samples == 0)
         continue;
       FontClassInfo& fcinfo = (*font_class_array_)(font_index, c);
       fcinfo.cloud_features.Init(feature_space_size);
       for (int s = 0; s < num_samples; ++s) {
         const TrainingSample* sample = GetSample(font_id, c, s);
         const GenericVector<int>& sample_features = sample->indexed_features();
         for (int i = 0; i < sample_features.size(); ++i)
           fcinfo.cloud_features.SetBit(sample_features[i]);
       }
     }
   }
 }

◆ ComputeClusterDistance()

float tesseract::TrainingSampleSet::ComputeClusterDistance	(	int	font_id1,
		int	class_id1,
		int	font_id2,
		int	class_id2,
		const IntFeatureMap &	feature_map
	)		const

Definition at line 367 of file trainingsampleset.cpp.

                                             {
   int dist = ReliablySeparable(font_id1, class_id1, font_id2, class_id2,
                                feature_map, false);
   dist += ReliablySeparable(font_id2, class_id2, font_id1, class_id1,
                             feature_map, false);
   int denominator = GetCanonicalFeatures(font_id1, class_id1).size();
   denominator += GetCanonicalFeatures(font_id2, class_id2).size();
   return static_cast<float>(dist) / denominator;
 }

◆ DeleteableSample()

bool tesseract::TrainingSampleSet::DeleteableSample ( const TrainingSample * sample )

Definition at line 506 of file trainingsampleset.cpp.

                                                                      {
   return sample == NULL || sample->class_id() < 0;
 }

◆ DeleteDeadSamples()

void tesseract::TrainingSampleSet::DeleteDeadSamples ( )

Definition at line 497 of file trainingsampleset.cpp.

                                           {
   samples_.compact(
       NewPermanentTessCallback(this, &TrainingSampleSet::DeleteableSample));
   num_raw_samples_ = samples_.size();
   // Samples must be re-organized now we have deleted a few.
 }

◆ DeSerialize()

bool tesseract::TrainingSampleSet::DeSerialize	(	bool	swap,
		FILE *	fp
	)

Definition at line 94 of file trainingsampleset.cpp.

                                                        {
   if (!samples_.DeSerialize(swap, fp)) return false;
   num_raw_samples_ = samples_.size();
   if (!unicharset_.load_from_file(fp)) return false;
   if (!font_id_map_.DeSerialize(swap, fp)) return false;
   delete font_class_array_;
   font_class_array_ = NULL;
   inT8 not_null;
   if (fread(&not_null, sizeof(not_null), 1, fp) != 1) return false;
   if (not_null) {
     FontClassInfo empty;
     font_class_array_ = new GENERIC_2D_ARRAY<FontClassInfo >(1, 1 , empty);
     if (!font_class_array_->DeSerializeClasses(swap, fp)) return false;
   }
   unicharset_size_ = unicharset_.size();
   return true;
 }

◆ DisplaySamplesWithFeature()

void tesseract::TrainingSampleSet::DisplaySamplesWithFeature	(	int	f_index,
		const Shape &	shape,
		const IntFeatureSpace &	feature_space,
		ScrollView::Color	color,
		ScrollView *	window
	)		const

Definition at line 743 of file trainingsampleset.cpp.

                                                                             {
   for (int s = 0; s < num_raw_samples(); ++s) {
     const TrainingSample* sample = GetSample(s);
     if (shape.ContainsUnichar(sample->class_id())) {
       GenericVector<int> indexed_features;
       space.IndexAndSortFeatures(sample->features(), sample->num_features(),
                                  &indexed_features);
       for (int f = 0; f < indexed_features.size(); ++f) {
         if (indexed_features[f] == f_index) {
           sample->DisplayFeatures(color, window);
         }
       }
     }
   }
 }

◆ extract_sample()

TrainingSample* tesseract::TrainingSampleSet::extract_sample ( int index )

inline

Definition at line 165 of file trainingsampleset.h.

                                             {
     TrainingSample* sample = samples_[index];
     samples_[index] = NULL;
     return sample;
   }

◆ fontinfo_table()

const FontInfoTable& tesseract::TrainingSampleSet::fontinfo_table ( ) const

inline

Definition at line 70 of file trainingsampleset.h.

                                               {
     return fontinfo_table_;
   }

◆ GetCanonicalDist()

float tesseract::TrainingSampleSet::GetCanonicalDist	(	int	font_id,
		int	class_id
	)		const

Definition at line 474 of file trainingsampleset.cpp.

                                                                          {
   ASSERT_HOST(font_class_array_ != NULL);
   int font_index = font_id_map_.SparseToCompact(font_id);
   if (font_index < 0) return 0.0f;
   if ((*font_class_array_)(font_index, class_id).canonical_sample >= 0)
     return (*font_class_array_)(font_index, class_id).canonical_dist;
   else
     return 0.0f;
 }

◆ GetCanonicalFeatures()

const GenericVector< int > & tesseract::TrainingSampleSet::GetCanonicalFeatures	(	int	font_id,
		int	class_id
	)		const

Definition at line 219 of file trainingsampleset.cpp.

                                      {
   int font_index = font_id_map_.SparseToCompact(font_id);
   ASSERT_HOST(font_index >= 0);
   return (*font_class_array_)(font_index, class_id).canonical_features;
 }

◆ GetCanonicalSample()

const TrainingSample * tesseract::TrainingSampleSet::GetCanonicalSample	(	int	font_id,
		int	class_id
	)		const

Definition at line 462 of file trainingsampleset.cpp.

                                      {
   ASSERT_HOST(font_class_array_ != NULL);
   int font_index = font_id_map_.SparseToCompact(font_id);
   if (font_index < 0) return NULL;
   int sample_index = (*font_class_array_)(font_index,
                                           class_id).canonical_sample;
   return sample_index >= 0 ? samples_[sample_index] : NULL;
 }

◆ GetCloudFeatures()

const BitVector & tesseract::TrainingSampleSet::GetCloudFeatures	(	int	font_id,
		int	class_id
	)		const

Definition at line 211 of file trainingsampleset.cpp.

                                      {
   int font_index = font_id_map_.SparseToCompact(font_id);
   ASSERT_HOST(font_index >= 0);
   return (*font_class_array_)(font_index, class_id).cloud_features;
 }

◆ GetSample() [1/2]

const TrainingSample * tesseract::TrainingSampleSet::GetSample ( int index ) const

Definition at line 174 of file trainingsampleset.cpp.

                                                                   {
   return samples_[index];
 }

◆ GetSample() [2/2]

const TrainingSample * tesseract::TrainingSampleSet::GetSample	(	int	font_id,
		int	class_id,
		int	index
	)		const

Definition at line 180 of file trainingsampleset.cpp.

                                                                     {
   ASSERT_HOST(font_class_array_ != NULL);
   int font_index = font_id_map_.SparseToCompact(font_id);
   if (font_index < 0) return NULL;
   int sample_index = (*font_class_array_)(font_index, class_id).samples[index];
   return samples_[sample_index];
 }

◆ GlobalSampleIndex()

int tesseract::TrainingSampleSet::GlobalSampleIndex	(	int	font_id,
		int	class_id,
		int	index
	)		const

Definition at line 452 of file trainingsampleset.cpp.

                                                           {
   ASSERT_HOST(font_class_array_ != NULL);
   int font_index = font_id_map_.SparseToCompact(font_id);
   if (font_index < 0) return -1;
   return (*font_class_array_)(font_index, class_id).samples[index];
 }

◆ IndexFeatures()

void tesseract::TrainingSampleSet::IndexFeatures ( const IntFeatureSpace & feature_space )

Definition at line 485 of file trainingsampleset.cpp.

                                                                           {
   for (int s = 0; s < samples_.size(); ++s)
     samples_[s]->IndexFeatures(feature_space);
 }

◆ KillSample()

void tesseract::TrainingSampleSet::KillSample ( TrainingSample * sample )

Definition at line 492 of file trainingsampleset.cpp.

                                                          {
   sample->set_sample_index(-1);
 }

◆ LoadUnicharset()

void tesseract::TrainingSampleSet::LoadUnicharset ( const char * filename )

Definition at line 113 of file trainingsampleset.cpp.

                                                            {
   if (!unicharset_.load_from_file(filename)) {
     tprintf("Failed to load unicharset from file %s\n"
             "Building unicharset from scratch...\n",
             filename);
     unicharset_.clear();
     // Add special characters as they were removed by the clear.
     UNICHARSET empty;
     unicharset_.AppendOtherUnicharset(empty);
   }
   unicharset_size_ = unicharset_.size();
 }

◆ mutable_sample()

TrainingSample* tesseract::TrainingSampleSet::mutable_sample ( int index )

inline

Definition at line 161 of file trainingsampleset.h.

                                             {
     return samples_[index];
   }

◆ MutableSample()

TrainingSample * tesseract::TrainingSampleSet::MutableSample	(	int	font_id,
		int	class_id,
		int	index
	)

Definition at line 191 of file trainingsampleset.cpp.

                                                             {
   ASSERT_HOST(font_class_array_ != NULL);
   int font_index = font_id_map_.SparseToCompact(font_id);
   if (font_index < 0) return NULL;
   int sample_index = (*font_class_array_)(font_index, class_id).samples[index];
   return samples_[sample_index];
 }

◆ num_raw_samples()

int tesseract::TrainingSampleSet::num_raw_samples ( ) const

inline

Definition at line 58 of file trainingsampleset.h.

                               {
     return num_raw_samples_;
   }

◆ num_samples()

int tesseract::TrainingSampleSet::num_samples ( ) const

inline

Definition at line 55 of file trainingsampleset.h.

                           {
     return samples_.size();
   }

◆ NumClassSamples()

int tesseract::TrainingSampleSet::NumClassSamples	(	int	font_id,
		int	class_id,
		bool	randomize
	)		const

Definition at line 156 of file trainingsampleset.cpp.

                                                              {
   ASSERT_HOST(font_class_array_ != NULL);
   if (font_id < 0 || class_id < 0 ||
       font_id >= font_id_map_.SparseSize() || class_id >= unicharset_size_) {
     // There are no samples because the font or class doesn't exist.
     return 0;
   }
   int font_index = font_id_map_.SparseToCompact(font_id);
   if (font_index < 0)
     return 0;  // The font has no samples.
   if (randomize)
     return (*font_class_array_)(font_index, class_id).samples.size();
   else
     return (*font_class_array_)(font_index, class_id).num_raw_samples;
 }

◆ NumFonts()

int tesseract::TrainingSampleSet::NumFonts ( ) const

inline

Definition at line 61 of file trainingsampleset.h.

                        {
     return font_id_map_.SparseSize();
   }

◆ OrganizeByFontAndClass()

void tesseract::TrainingSampleSet::OrganizeByFontAndClass ( )

Definition at line 511 of file trainingsampleset.cpp.

                                                {
   // Font indexes are sparse, so we used a map to compact them, so we can
   // have an efficient 2-d array of fonts and character classes.
   SetupFontIdMap();
   int compact_font_size = font_id_map_.CompactSize();
   // Get a 2-d array of generic vectors.
   if (font_class_array_ != NULL)
     delete font_class_array_;
   FontClassInfo empty;
   font_class_array_ = new GENERIC_2D_ARRAY<FontClassInfo>(
       compact_font_size, unicharset_size_, empty);
   for (int s = 0; s < samples_.size(); ++s) {
     int font_id = samples_[s]->font_id();
     int class_id = samples_[s]->class_id();
     if (font_id < 0 || font_id >= font_id_map_.SparseSize()) {
       tprintf("Font id = %d/%d, class id = %d/%d on sample %d\n",
               font_id, font_id_map_.SparseSize(), class_id, unicharset_size_,
               s);
     }
     ASSERT_HOST(font_id >= 0 && font_id < font_id_map_.SparseSize());
     ASSERT_HOST(class_id >= 0 && class_id < unicharset_size_);
     int font_index = font_id_map_.SparseToCompact(font_id);
     (*font_class_array_)(font_index, class_id).samples.push_back(s);
   }
   // Set the num_raw_samples member of the FontClassInfo, to set the boundary
   // between the raw samples and the replicated ones.
   for (int f = 0; f < compact_font_size; ++f) {
     for (int c = 0; c < unicharset_size_; ++c)
       (*font_class_array_)(f, c).num_raw_samples =
           (*font_class_array_)(f, c).samples.size();
   }
   // This is the global number of samples and also marks the boundary between
   // real and replicated samples.
   num_raw_samples_ = samples_.size();
 }

◆ ReliablySeparable()

int tesseract::TrainingSampleSet::ReliablySeparable	(	int	font_id1,
		int	class_id1,
		int	font_id2,
		int	class_id2,
		const IntFeatureMap &	feature_map,
		bool	thorough
	)		const

Definition at line 413 of file trainingsampleset.cpp.

                                                               {
   int result = 0;
   const TrainingSample* sample2 = GetCanonicalSample(font_id2, class_id2);
   if (sample2 == NULL)
     return 0;  // There are no canonical features.
   const GenericVector<int>& canonical2 = GetCanonicalFeatures(font_id2,
                                                               class_id2);
   const BitVector& cloud1 = GetCloudFeatures(font_id1, class_id1);
   if (cloud1.size() == 0)
     return canonical2.size();  // There are no cloud features.
 
   // Find a canonical2 feature that is not in cloud1.
   for (int f = 0; f < canonical2.size(); ++f) {
     int feature = canonical2[f];
     if (cloud1[feature])
       continue;
     // Gather the near neighbours of f.
     GenericVector<int> good_features;
     AddNearFeatures(feature_map, feature, 1, &good_features);
     // Check that none of the good_features are in the cloud.
     int i;
     for (i = 0; i < good_features.size(); ++i) {
       int good_f = good_features[i];
       if (cloud1[good_f]) {
         break;
       }
     }
     if (i < good_features.size())
        continue;  // Found one in the cloud.
     ++result;
   }
   return result;
 }

◆ ReplicateAndRandomizeSamples()

void tesseract::TrainingSampleSet::ReplicateAndRandomizeSamples ( )

Definition at line 665 of file trainingsampleset.cpp.

                                                      {
   ASSERT_HOST(font_class_array_ != NULL);
   int font_size = font_id_map_.CompactSize();
   for (int font_index = 0; font_index < font_size; ++font_index) {
     for (int c = 0; c < unicharset_size_; ++c) {
       FontClassInfo& fcinfo = (*font_class_array_)(font_index, c);
       int sample_count = fcinfo.samples.size();
       int min_samples = 2 * MAX(kSampleRandomSize, sample_count);
       if (sample_count > 0 && sample_count < min_samples) {
         int base_count = sample_count;
         for (int base_index = 0; sample_count < min_samples; ++sample_count) {
           int src_index = fcinfo.samples[base_index++];
           if (base_index >= base_count) base_index = 0;
           TrainingSample* sample = samples_[src_index]->RandomizedCopy(
               sample_count % kSampleRandomSize);
           int sample_index = samples_.size();
           sample->set_sample_index(sample_index);
           samples_.push_back(sample);
           fcinfo.samples.push_back(sample_index);
         }
       }
     }
   }
 }

◆ SampleToString()

STRING tesseract::TrainingSampleSet::SampleToString ( const TrainingSample & sample ) const

Definition at line 202 of file trainingsampleset.cpp.

                                                                            {
   STRING boxfile_str;
   MakeBoxFileStr(unicharset_.id_to_unichar(sample.class_id()),
                  sample.bounding_box(), sample.page_num(), &boxfile_str);
   return STRING(fontinfo_table_.get(sample.font_id()).name) + " " + boxfile_str;
 }

◆ Serialize()

bool tesseract::TrainingSampleSet::Serialize ( FILE * fp ) const

Definition at line 80 of file trainingsampleset.cpp.

                                                 {
   if (!samples_.Serialize(fp)) return false;
   if (!unicharset_.save_to_file(fp)) return false;
   if (!font_id_map_.Serialize(fp)) return false;
   inT8 not_null = font_class_array_ != NULL;
   if (fwrite(&not_null, sizeof(not_null), 1, fp) != 1) return false;
   if (not_null) {
     if (!font_class_array_->SerializeClasses(fp)) return false;
   }
   return true;
 }

◆ SetupFontIdMap()

void tesseract::TrainingSampleSet::SetupFontIdMap ( )

Definition at line 549 of file trainingsampleset.cpp.

                                        {
   // Number of samples for each font_id.
   GenericVector<int> font_counts;
   for (int s = 0; s < samples_.size(); ++s) {
     int font_id = samples_[s]->font_id();
     while (font_id >= font_counts.size())
       font_counts.push_back(0);
     ++font_counts[font_id];
   }
   font_id_map_.Init(font_counts.size(), false);
   for (int f = 0; f < font_counts.size(); ++f) {
     font_id_map_.SetMap(f, font_counts[f] > 0);
   }
   font_id_map_.Setup();
 }

◆ UnicharDistance()

float tesseract::TrainingSampleSet::UnicharDistance	(	const UnicharAndFonts &	uf1,
		const UnicharAndFonts &	uf2,
		bool	matched_fonts,
		const IntFeatureMap &	feature_map
	)

Definition at line 230 of file trainingsampleset.cpp.

                                                                            {
   int num_fonts1 = uf1.font_ids.size();
   int c1 = uf1.unichar_id;
   int num_fonts2 = uf2.font_ids.size();
   int c2 = uf2.unichar_id;
   double dist_sum = 0.0;
   int dist_count = 0;
   bool debug = false;
   if (matched_fonts) {
     // Compute distances only where fonts match.
     for (int i = 0; i < num_fonts1; ++i) {
       int f1 = uf1.font_ids[i];
       for (int j = 0; j < num_fonts2; ++j) {
         int f2 = uf2.font_ids[j];
         if (f1 == f2) {
           dist_sum += ClusterDistance(f1, c1, f2, c2, feature_map);
           ++dist_count;
         }
       }
     }
   } else if (num_fonts1 * num_fonts2 <= kSquareLimit) {
     // Small enough sets to compute all the distances.
     for (int i = 0; i < num_fonts1; ++i) {
       int f1 = uf1.font_ids[i];
       for (int j = 0; j < num_fonts2; ++j) {
         int f2 = uf2.font_ids[j];
         dist_sum += ClusterDistance(f1, c1, f2, c2, feature_map);
         if (debug) {
             tprintf("Cluster dist %d %d %d %d = %g\n",
                     f1, c1, f2, c2,
                     ClusterDistance(f1, c1, f2, c2, feature_map));
         }
         ++dist_count;
       }
     }
   } else {
     // Subsample distances, using the largest set once, and stepping through
     // the smaller set so as to ensure that all the pairs are different.
     int increment = kPrime1 != num_fonts2 ? kPrime1 : kPrime2;
     int index = 0;
     int num_samples = MAX(num_fonts1, num_fonts2);
     for (int i = 0; i < num_samples; ++i, index += increment) {
       int f1 = uf1.font_ids[i % num_fonts1];
       int f2 = uf2.font_ids[index % num_fonts2];
       if (debug) {
           tprintf("Cluster dist %d %d %d %d = %g\n",
                   f1, c1, f2, c2, ClusterDistance(f1, c1, f2, c2, feature_map));
       }
       dist_sum += ClusterDistance(f1, c1, f2, c2, feature_map);
       ++dist_count;
     }
   }
   if (dist_count == 0) {
     if (matched_fonts)
       return UnicharDistance(uf1, uf2, false, feature_map);
     return 0.0f;
   }
   return dist_sum / dist_count;
 }

◆ unicharset()

const UNICHARSET& tesseract::TrainingSampleSet::unicharset ( ) const

inline

Definition at line 64 of file trainingsampleset.h.

                                        {
     return unicharset_;
   }

The documentation for this class was generated from the following files:

/home/stefan/src/github/tesseract-ocr/tesseract/classify/trainingsampleset.h
/home/stefan/src/github/tesseract-ocr/tesseract/classify/trainingsampleset.cpp

Public Member Functions

Detailed Description

Constructor & Destructor Documentation

◆ TrainingSampleSet()

◆ ~TrainingSampleSet()

Member Function Documentation

◆ AddAllFontsForClass()

◆ AddSample() [1/2]

◆ AddSample() [2/2]

◆ charsetsize()

◆ ClusterDistance()

◆ ComputeCanonicalFeatures()

◆ ComputeCanonicalSamples()

◆ ComputeCloudFeatures()

◆ ComputeClusterDistance()

◆ DeleteableSample()

◆ DeleteDeadSamples()

◆ DeSerialize()

◆ DisplaySamplesWithFeature()

◆ extract_sample()

◆ fontinfo_table()

◆ GetCanonicalDist()

◆ GetCanonicalFeatures()

◆ GetCanonicalSample()

◆ GetCloudFeatures()

◆ GetSample() [1/2]

◆ GetSample() [2/2]

◆ GlobalSampleIndex()

◆ IndexFeatures()

◆ KillSample()

◆ LoadUnicharset()

◆ mutable_sample()

◆ MutableSample()

◆ num_raw_samples()

◆ num_samples()

◆ NumClassSamples()

◆ NumFonts()

◆ OrganizeByFontAndClass()

◆ ReliablySeparable()

◆ ReplicateAndRandomizeSamples()

◆ SampleToString()

◆ Serialize()

◆ SetupFontIdMap()

◆ UnicharDistance()

◆ unicharset()