tesseract  5.0.0
cntraining.cpp File Reference
#include <tesseract/unichar.h>
#include <cmath>
#include <cstdio>
#include <cstring>
#include "cluster.h"
#include "clusttool.h"
#include "commontraining.h"
#include "featdefs.h"
#include "ocrfeatures.h"
#include "oldlist.h"

Go to the source code of this file.

Macros

#define PROGRAM_FEATURE_TYPE   "cn"
 

Functions

int main (int argc, char *argv[])
 

Macro Definition Documentation

◆ PROGRAM_FEATURE_TYPE

#define PROGRAM_FEATURE_TYPE   "cn"

Definition at line 33 of file cntraining.cpp.

Function Documentation

◆ main()

int main ( int  argc,
char *  argv[] 
)

This program reads in a text file consisting of feature samples from a training page in the following format:

   FontName CharName NumberOfFeatureTypes(N)
      FeatureTypeName1 NumberOfFeatures(M)
         Feature1
         ...
         FeatureM
      FeatureTypeName2 NumberOfFeatures(M)
         Feature1
         ...
         FeatureM
      ...
      FeatureTypeNameN NumberOfFeatures(M)
         Feature1
         ...
         FeatureM
   FontName CharName ...

It then appends these samples into a separate file for each character. The name of the file is

DirectoryName/FontName/CharName.FeatureTypeName

The DirectoryName can be specified via a command line argument. If not specified, it defaults to the current directory. The format of the resulting files is:

   NumberOfFeatures(M)
      Feature1
      ...
      FeatureM
   NumberOfFeatures(M)
   ...

The output files each have a header which describes the type of feature which the file contains. This header is in the format required by the clusterer. A command line argument can also be used to specify that only the first N samples of each class should be used.

Parameters
argcnumber of command line arguments
argvarray of command line arguments
Returns
0 on success

Definition at line 103 of file cntraining.cpp.

103  {
104  tesseract::CheckSharedLibraryVersion();
105 
106  // Set the global Config parameters before parsing the command line.
107  Config = CNConfig;
108 
109  LIST CharList = NIL_LIST;
110  CLUSTERER *Clusterer = nullptr;
111  LIST ProtoList = NIL_LIST;
112  LIST NormProtoList = NIL_LIST;
113  LIST pCharList;
114  LABELEDLIST CharSample;
115  FEATURE_DEFS_STRUCT FeatureDefs;
116  InitFeatureDefs(&FeatureDefs);
117 
118  ParseArguments(&argc, &argv);
119  int num_fonts = 0;
120  for (const char *PageName = *++argv; PageName != nullptr; PageName = *++argv) {
121  printf("Reading %s ...\n", PageName);
122  FILE *TrainingPage = fopen(PageName, "rb");
123  ASSERT_HOST(TrainingPage);
124  if (TrainingPage) {
125  ReadTrainingSamples(FeatureDefs, PROGRAM_FEATURE_TYPE, 100, nullptr, TrainingPage, &CharList);
126  fclose(TrainingPage);
127  ++num_fonts;
128  }
129  }
130  printf("Clustering ...\n");
131  // To allow an individual font to form a separate cluster,
132  // reduce the min samples:
133  // Config.MinSamples = 0.5 / num_fonts;
134  pCharList = CharList;
135  // The norm protos will count the source protos, so we keep them here in
136  // freeable_protos, so they can be freed later.
137  std::vector<LIST> freeable_protos;
138  iterate(pCharList) {
139  // Cluster
140  CharSample = reinterpret_cast<LABELEDLIST>(pCharList->first_node());
141  Clusterer = SetUpForClustering(FeatureDefs, CharSample, PROGRAM_FEATURE_TYPE);
142  if (Clusterer == nullptr) { // To avoid a SIGSEGV
143  fprintf(stderr, "Error: nullptr clusterer!\n");
144  return 1;
145  }
146  float SavedMinSamples = Config.MinSamples;
147  // To disable the tendency to produce a single cluster for all fonts,
148  // make MagicSamples an impossible to achieve number:
149  // Config.MagicSamples = CharSample->SampleCount * 10;
150  Config.MagicSamples = CharSample->SampleCount;
151  while (Config.MinSamples > 0.001) {
152  ProtoList = ClusterSamples(Clusterer, &Config);
153  if (NumberOfProtos(ProtoList, true, false) > 0) {
154  break;
155  } else {
156  Config.MinSamples *= 0.95;
157  printf(
158  "0 significant protos for %s."
159  " Retrying clustering with MinSamples = %f%%\n",
160  CharSample->Label.c_str(), Config.MinSamples);
161  }
162  }
163  Config.MinSamples = SavedMinSamples;
164  AddToNormProtosList(&NormProtoList, ProtoList, CharSample->Label);
165  freeable_protos.push_back(ProtoList);
166  FreeClusterer(Clusterer);
167  }
168  FreeTrainingSamples(CharList);
169  int desc_index = ShortNameToFeatureType(FeatureDefs, PROGRAM_FEATURE_TYPE);
170  WriteNormProtos(FLAGS_D.c_str(), NormProtoList, FeatureDefs.FeatureDesc[desc_index]);
171  FreeNormProtoList(NormProtoList);
172  for (auto &freeable_proto : freeable_protos) {
173  FreeProtoList(&freeable_proto);
174  }
175  printf("\n");
176  return 0;
177 } // main
#define ASSERT_HOST(x)
Definition: errcode.h:59
#define iterate(l)
Definition: oldlist.h:91
#define NIL_LIST
Definition: oldlist.h:75
#define PROGRAM_FEATURE_TYPE
Definition: cntraining.cpp:33
uint32_t ShortNameToFeatureType(const FEATURE_DEFS_STRUCT &FeatureDefs, const char *ShortName)
Definition: featdefs.cpp:203
void ReadTrainingSamples(const FEATURE_DEFS_STRUCT &feature_definitions, const char *feature_name, int max_samples, UNICHARSET *unicharset, FILE *file, LIST *training_samples)
void ParseArguments(int *argc, char ***argv)
void FreeNormProtoList(LIST CharList)
CLUSTERER * SetUpForClustering(const FEATURE_DEFS_STRUCT &FeatureDefs, LABELEDLIST char_sample, const char *program_feature_type)
CLUSTERCONFIG Config
void AddToNormProtosList(LIST *NormProtoList, LIST ProtoList, const std::string &CharName)
void InitFeatureDefs(FEATURE_DEFS_STRUCT *featuredefs)
Definition: featdefs.cpp:87
void FreeProtoList(LIST *ProtoList)
Definition: cluster.cpp:1598
void FreeTrainingSamples(LIST CharList)
void FreeClusterer(CLUSTERER *Clusterer)
Definition: cluster.cpp:1576
int NumberOfProtos(LIST ProtoList, bool CountSigProtos, bool CountInsigProtos)
LIST ClusterSamples(CLUSTERER *Clusterer, CLUSTERCONFIG *Config)
Definition: cluster.cpp:1544
const FEATURE_DESC_STRUCT * FeatureDesc[NUM_FEATURE_TYPES]
Definition: featdefs.h:43
list_rec * first_node()
Definition: oldlist.h:107