tesseract  5.0.0-alpha-619-ge9db
tesseract::LMPainPoints Class Reference

#include <lm_pain_points.h>

Public Member Functions

 LMPainPoints (int max, float rat, bool fp, const Dict *d, int deb)
 
 ~LMPainPoints ()
 
bool HasPainPoints (LMPainPointsType pp_type) const
 
LMPainPointsType Deque (MATRIX_COORD *pp, float *priority)
 
void Clear ()
 
void GenerateInitial (WERD_RES *word_res)
 
void GenerateFromPath (float rating_cert_scale, ViterbiStateEntry *vse, WERD_RES *word_res)
 
void GenerateFromAmbigs (const DANGERR &fixpt, ViterbiStateEntry *vse, WERD_RES *word_res)
 
bool GeneratePainPoint (int col, int row, LMPainPointsType pp_type, float special_priority, bool ok_to_extend, float max_char_wh_ratio, WERD_RES *word_res)
 
void RemapForSplit (int index)
 

Static Public Member Functions

static const char * PainPointDescription (LMPainPointsType type)
 

Static Public Attributes

static const float kDefaultPainPointPriorityAdjustment = 2.0f
 
static const float kLooseMaxCharWhRatio = 2.5f
 

Detailed Description

Definition at line 56 of file lm_pain_points.h.

Constructor & Destructor Documentation

◆ LMPainPoints()

tesseract::LMPainPoints::LMPainPoints ( int  max,
float  rat,
bool  fp,
const Dict d,
int  deb 
)
inline

Definition at line 69 of file lm_pain_points.h.

69  :
70  max_heap_size_(max), max_char_wh_ratio_(rat), fixed_pitch_(fp),
71  dict_(d), debug_level_(deb) {}

◆ ~LMPainPoints()

tesseract::LMPainPoints::~LMPainPoints ( )
inline

Definition at line 72 of file lm_pain_points.h.

72 {}

Member Function Documentation

◆ Clear()

void tesseract::LMPainPoints::Clear ( )
inline

Definition at line 85 of file lm_pain_points.h.

85  {
86  for (auto & pain_points_heap : pain_points_heaps_) pain_points_heap.clear();
87  }

◆ Deque()

LMPainPointsType tesseract::LMPainPoints::Deque ( MATRIX_COORD pp,
float *  priority 
)

Definition at line 39 of file lm_pain_points.cpp.

39  {
40  for (int h = 0; h < LM_PPTYPE_NUM; ++h) {
41  if (pain_points_heaps_[h].empty()) continue;
42  *priority = pain_points_heaps_[h].PeekTop().key;
43  *pp = pain_points_heaps_[h].PeekTop().data;
44  pain_points_heaps_[h].Pop(nullptr);
45  return static_cast<LMPainPointsType>(h);
46  }
47  return LM_PPTYPE_NUM;
48 }

◆ GenerateFromAmbigs()

void tesseract::LMPainPoints::GenerateFromAmbigs ( const DANGERR fixpt,
ViterbiStateEntry vse,
WERD_RES word_res 
)

Definition at line 132 of file lm_pain_points.cpp.

134  {
135  // Begins and ends in DANGERR vector now record the blob indices as used
136  // by the ratings matrix.
137  for (int d = 0; d < fixpt.size(); ++d) {
138  const DANGERR_INFO &danger = fixpt[d];
139  // Only use dangerous ambiguities.
140  if (danger.dangerous) {
141  GeneratePainPoint(danger.begin, danger.end - 1,
142  LM_PPTYPE_AMBIG, vse->cost, true,
143  kLooseMaxCharWhRatio, word_res);
144  }
145  }
146 }

◆ GenerateFromPath()

void tesseract::LMPainPoints::GenerateFromPath ( float  rating_cert_scale,
ViterbiStateEntry vse,
WERD_RES word_res 
)

Definition at line 70 of file lm_pain_points.cpp.

72  {
73  ViterbiStateEntry *curr_vse = vse;
74  BLOB_CHOICE *curr_b = vse->curr_b;
75  // The following pain point generation and priority calculation approaches
76  // prioritize exploring paths with low average rating of the known part of
77  // the path, while not relying on the ratings of the pieces to be combined.
78  //
79  // A pain point to combine the neighbors is generated for each pair of
80  // neighboring blobs on the path (the path is represented by vse argument
81  // given to GenerateFromPath()). The priority of each pain point is set to
82  // the average rating (per outline length) of the path, not including the
83  // ratings of the blobs to be combined.
84  // The ratings of the blobs to be combined are not used to calculate the
85  // priority, since it is not possible to determine from their magnitude
86  // whether it will be beneficial to combine the blobs. The reason is that
87  // chopped junk blobs (/ | - ') can have very good (low) ratings, however
88  // combining them will be beneficial. Blobs with high ratings might be
89  // over-joined pieces of characters, but also could be blobs from an unseen
90  // font or chopped pieces of complex characters.
91  while (curr_vse->parent_vse != nullptr) {
92  ViterbiStateEntry* parent_vse = curr_vse->parent_vse;
93  const MATRIX_COORD& curr_cell = curr_b->matrix_cell();
94  const MATRIX_COORD& parent_cell = parent_vse->curr_b->matrix_cell();
95  MATRIX_COORD pain_coord(parent_cell.col, curr_cell.row);
96  if (!pain_coord.Valid(*word_res->ratings) ||
97  !word_res->ratings->Classified(parent_cell.col, curr_cell.row,
98  dict_->WildcardID())) {
99  // rat_subtr contains ratings sum of the two adjacent blobs to be merged.
100  // rat_subtr will be subtracted from the ratings sum of the path, since
101  // the blobs will be joined into a new blob, whose rating is yet unknown.
102  float rat_subtr = curr_b->rating() + parent_vse->curr_b->rating();
103  // ol_subtr contains the outline length of the blobs that will be joined.
104  float ol_subtr =
105  AssociateUtils::ComputeOutlineLength(rating_cert_scale, *curr_b) +
106  AssociateUtils::ComputeOutlineLength(rating_cert_scale,
107  *(parent_vse->curr_b));
108  // ol_dif is the outline of the path without the two blobs to be joined.
109  float ol_dif = vse->outline_length - ol_subtr;
110  // priority is set to the average rating of the path per unit of outline,
111  // not counting the ratings of the pieces to be joined.
112  float priority = ol_dif > 0 ? (vse->ratings_sum-rat_subtr)/ol_dif : 0.0;
113  GeneratePainPoint(pain_coord.col, pain_coord.row, LM_PPTYPE_PATH,
114  priority, true, max_char_wh_ratio_, word_res);
115  } else if (debug_level_ > 3) {
116  tprintf("NO pain point (Classified) for col=%d row=%d type=%s\n",
117  pain_coord.col, pain_coord.row,
118  LMPainPointsTypeName[LM_PPTYPE_PATH]);
119  BLOB_CHOICE_IT b_it(word_res->ratings->get(pain_coord.col,
120  pain_coord.row));
121  for (b_it.mark_cycle_pt(); !b_it.cycled_list(); b_it.forward()) {
122  BLOB_CHOICE* choice = b_it.data();
123  choice->print_full();
124  }
125  }
126 
127  curr_vse = parent_vse;
128  curr_b = curr_vse->curr_b;
129  }
130 }

◆ GenerateInitial()

void tesseract::LMPainPoints::GenerateInitial ( WERD_RES word_res)

Definition at line 50 of file lm_pain_points.cpp.

50  {
51  MATRIX *ratings = word_res->ratings;
52  AssociateStats associate_stats;
53  for (int col = 0; col < ratings->dimension(); ++col) {
54  int row_end = std::min(ratings->dimension(), col + ratings->bandwidth() + 1);
55  for (int row = col + 1; row < row_end; ++row) {
56  MATRIX_COORD coord(col, row);
57  if (coord.Valid(*ratings) &&
58  ratings->get(col, row) != NOT_CLASSIFIED) continue;
59  // Add an initial pain point if needed.
60  if (ratings->Classified(col, row - 1, dict_->WildcardID()) ||
61  (col + 1 < ratings->dimension() &&
62  ratings->Classified(col + 1, row, dict_->WildcardID()))) {
63  GeneratePainPoint(col, row, LM_PPTYPE_SHAPE, 0.0,
64  true, max_char_wh_ratio_, word_res);
65  }
66  }
67  }
68 }

◆ GeneratePainPoint()

bool tesseract::LMPainPoints::GeneratePainPoint ( int  col,
int  row,
LMPainPointsType  pp_type,
float  special_priority,
bool  ok_to_extend,
float  max_char_wh_ratio,
WERD_RES word_res 
)

Definition at line 148 of file lm_pain_points.cpp.

151  {
152  MATRIX_COORD coord(col, row);
153  if (coord.Valid(*word_res->ratings) &&
154  word_res->ratings->Classified(col, row, dict_->WildcardID())) {
155  return false;
156  }
157  if (debug_level_ > 3) {
158  tprintf("Generating pain point for col=%d row=%d type=%s\n",
159  col, row, LMPainPointsTypeName[pp_type]);
160  }
161  // Compute associate stats.
162  AssociateStats associate_stats;
163  AssociateUtils::ComputeStats(col, row, nullptr, 0, fixed_pitch_,
164  max_char_wh_ratio, word_res, debug_level_,
165  &associate_stats);
166  // For fixed-pitch fonts/languages: if the current combined blob overlaps
167  // the next blob on the right and it is ok to extend the blob, try extending
168  // the blob until there is no overlap with the next blob on the right or
169  // until the width-to-height ratio becomes too large.
170  if (ok_to_extend) {
171  while (associate_stats.bad_fixed_pitch_right_gap &&
172  row + 1 < word_res->ratings->dimension() &&
173  !associate_stats.bad_fixed_pitch_wh_ratio) {
174  AssociateUtils::ComputeStats(col, ++row, nullptr, 0, fixed_pitch_,
175  max_char_wh_ratio, word_res, debug_level_,
176  &associate_stats);
177  }
178  }
179  if (associate_stats.bad_shape) {
180  if (debug_level_ > 3) {
181  tprintf("Discarded pain point with a bad shape\n");
182  }
183  return false;
184  }
185 
186  // Insert the new pain point into pain_points_heap_.
187  if (pain_points_heaps_[pp_type].size() < max_heap_size_) {
188  // Compute pain point priority.
189  float priority;
190  if (pp_type == LM_PPTYPE_PATH) {
191  priority = special_priority;
192  } else {
193  priority = associate_stats.gap_sum;
194  }
195  MatrixCoordPair pain_point(priority, MATRIX_COORD(col, row));
196  pain_points_heaps_[pp_type].Push(&pain_point);
197  if (debug_level_) {
198  tprintf("Added pain point with priority %g\n", priority);
199  }
200  return true;
201  } else {
202  if (debug_level_) tprintf("Pain points heap is full\n");
203  return false;
204  }
205 }

◆ HasPainPoints()

bool tesseract::LMPainPoints::HasPainPoints ( LMPainPointsType  pp_type) const
inline

Definition at line 75 of file lm_pain_points.h.

75  {
76  return !pain_points_heaps_[pp_type].empty();
77  }

◆ PainPointDescription()

static const char* tesseract::LMPainPoints::PainPointDescription ( LMPainPointsType  type)
inlinestatic

Definition at line 65 of file lm_pain_points.h.

65  {
66  return LMPainPointsTypeName[type];
67  }

◆ RemapForSplit()

void tesseract::LMPainPoints::RemapForSplit ( int  index)

Adjusts the pain point coordinates to cope with expansion of the ratings matrix due to a split of the blob with the given index.

Definition at line 211 of file lm_pain_points.cpp.

211  {
212  for (auto & pain_points_heap : pain_points_heaps_) {
213  GenericVector<MatrixCoordPair>* heap = pain_points_heap.heap();
214  for (int j = 0; j < heap->size(); ++j)
215  (*heap)[j].data.MapForSplit(index);
216  }
217 }

Member Data Documentation

◆ kDefaultPainPointPriorityAdjustment

const float tesseract::LMPainPoints::kDefaultPainPointPriorityAdjustment = 2.0f
static

Definition at line 59 of file lm_pain_points.h.

◆ kLooseMaxCharWhRatio

const float tesseract::LMPainPoints::kLooseMaxCharWhRatio = 2.5f
static

Definition at line 63 of file lm_pain_points.h.


The documentation for this class was generated from the following files:
tesseract::GenericHeap::Pop
bool Pop(Pair *entry)
Definition: genericheap.h:118
tesseract::GenericHeap::PeekTop
const Pair & PeekTop() const
Definition: genericheap.h:108
MATRIX
Definition: matrix.h:574
DANGERR_INFO
Definition: stopper.h:33
DANGERR_INFO::dangerous
bool dangerous
Definition: stopper.h:41
WERD_RES::ratings
MATRIX * ratings
Definition: pageres.h:231
tesseract::KDPair::key
Key key
Definition: kdpair.h:46
BLOB_CHOICE::matrix_cell
const MATRIX_COORD & matrix_cell()
Definition: ratngs.h:115
tesseract::LMPainPoints::kLooseMaxCharWhRatio
static const float kLooseMaxCharWhRatio
Definition: lm_pain_points.h:63
NOT_CLASSIFIED
#define NOT_CLASSIFIED
Definition: matrix.h:40
tesseract::Dict::WildcardID
UNICHAR_ID WildcardID() const
Definition: dict.h:428
DANGERR_INFO::begin
int begin
Definition: stopper.h:39
tesseract::LM_PPTYPE_NUM
Definition: lm_pain_points.h:46
tesseract::LM_PPTYPE_PATH
Definition: lm_pain_points.h:43
GENERIC_2D_ARRAY::get
T get(ICOORD pos) const
Definition: matrix.h:227
tesseract::KDPair::data
Data data
Definition: kdpair.h:45
tesseract::LMPainPoints::GeneratePainPoint
bool GeneratePainPoint(int col, int row, LMPainPointsType pp_type, float special_priority, bool ok_to_extend, float max_char_wh_ratio, WERD_RES *word_res)
Definition: lm_pain_points.cpp:148
tesseract::LM_PPTYPE_SHAPE
Definition: lm_pain_points.h:44
BLOB_CHOICE::rating
float rating() const
Definition: ratngs.h:78
GenericVector< MatrixCoordPair >
tesseract::KDPairInc
Definition: kdpair.h:51
tesseract::GenericHeap::Push
void Push(Pair *entry)
Definition: genericheap.h:95
tesseract::AssociateUtils::ComputeOutlineLength
static float ComputeOutlineLength(float rating_cert_scale, const BLOB_CHOICE &b)
Definition: associate.h:80
BandTriMatrix::bandwidth
int bandwidth() const
Definition: matrix.h:534
BLOB_CHOICE
Definition: ratngs.h:49
MATRIX_COORD
Definition: matrix.h:604
MATRIX_COORD::col
int col
Definition: matrix.h:632
BLOB_CHOICE::print_full
void print_full() const
Definition: ratngs.h:175
tprintf
DLLSYM void tprintf(const char *format,...)
Definition: tprintf.cpp:34
tesseract::AssociateUtils::ComputeStats
static void ComputeStats(int col, int row, const AssociateStats *parent_stats, int parent_path_length, bool fixed_pitch, float max_char_wh_ratio, WERD_RES *word_res, bool debug, AssociateStats *stats)
Definition: associate.cpp:34
tesstrain_utils.type
type
Definition: tesstrain_utils.py:141
MATRIX_COORD::row
int row
Definition: matrix.h:633
tesseract::LM_PPTYPE_AMBIG
Definition: lm_pain_points.h:42
GenericVector::size
int size() const
Definition: genericvector.h:71
tesseract::GenericHeap::empty
bool empty() const
Definition: genericheap.h:68
DANGERR_INFO::end
int end
Definition: stopper.h:40
BandTriMatrix::dimension
int dimension() const
Definition: matrix.h:532
MATRIX::Classified
bool Classified(int col, int row, int wildcard_id) const
Definition: matrix.cpp:34