« back to all changes in this revision
Viewing changes to classify/classify.h
- browse files at revision 691
- compare with another revision
- download diff
- download tarball
- view history from revision 691
- Committer: theraysmith at gmail
- Date: 2013-11-08 20:30:56 UTC
- Revision ID: svn-v4:d0cd1f9f-072b-0410-8dd7-cf729c803f20:trunk:904
Refactorerd control functions to enable parallel blob classification
- files modified:
- ccmain/Makefile.am
added
removed
classify/classify.h
145
145
int FontinfoId,
146
146
ADAPT_CLASS Class,
147
147
ADAPT_TEMPLATES Templates);
148
void AdaptToPunc(TBLOB *Blob,
149
CLASS_ID ClassId,
150
int FontinfoId,
151
FLOAT32 Threshold);
152
void AmbigClassifier(TBLOB *Blob,
153
INT_TEMPLATES Templates,
154
ADAPT_CLASS *Classes,
155
UNICHAR_ID *Ambiguities,
156
ADAPT_RESULTS *Results);
148
void AmbigClassifier(const GenericVector<INT_FEATURE_STRUCT>& int_features,
149
const INT_FX_RESULT_STRUCT& fx_info,
150
const TBLOB *blob,
151
INT_TEMPLATES templates,
152
ADAPT_CLASS *classes,
153
UNICHAR_ID *ambiguities,
154
ADAPT_RESULTS *results);
157
155
void MasterMatcher(INT_TEMPLATES templates,
158
156
inT16 num_features,
159
157
const INT_FEATURE_STRUCT* features,
161
159
ADAPT_CLASS* classes,
162
160
int debug,
163
161
int num_classes,
162
int matcher_multiplier,
164
163
const TBOX& blob_box,
165
164
CLASS_PRUNER_RESULTS results,
166
165
ADAPT_RESULTS* final_results);
175
174
int bottom, int top,
176
175
float cp_rating,
177
176
int blob_length,
177
int matcher_multiplier,
178
178
const uinT8* cn_factors,
179
179
INT_RESULT_STRUCT& int_result,
180
180
ADAPT_RESULTS* final_results);
184
184
double ComputeCorrectedRating(bool debug, int unichar_id, double cp_rating,
185
185
double im_rating, int feature_misses,
186
186
int bottom, int top,
187
int blob_length, const uinT8* cn_factors);
187
int blob_length, int matcher_multiplier,
188
const uinT8* cn_factors);
188
189
void ConvertMatchesToChoices(const DENORM& denorm, const TBOX& box,
189
190
ADAPT_RESULTS *Results,
190
191
BLOB_CHOICE_LIST *Choices);
246
247
// Converts a shape_table_ index to a classifier class_id index (not a
247
248
// unichar-id!). Uses a search, so not fast.
248
249
int ShapeIDToClassID(int shape_id) const;
249
UNICHAR_ID *BaselineClassifier(TBLOB *Blob,
250
ADAPT_TEMPLATES Templates,
251
ADAPT_RESULTS *Results);
252
int CharNormClassifier(TBLOB *Blob,
253
INT_TEMPLATES Templates,
254
ADAPT_RESULTS *Results);
250
UNICHAR_ID *BaselineClassifier(
251
TBLOB *Blob, const GenericVector<INT_FEATURE_STRUCT>& int_features,
252
const INT_FX_RESULT_STRUCT& fx_info,
253
ADAPT_TEMPLATES Templates, ADAPT_RESULTS *Results);
254
int CharNormClassifier(TBLOB *blob,
255
const TrainingSample& sample,
256
ADAPT_RESULTS *adapt_results);
255
257
256
258
// As CharNormClassifier, but operates on a TrainingSample and outputs to
257
259
// a GenericVector of ShapeRating without conversion to classes.
267
269
void DisplayAdaptedChar(TBLOB* blob, INT_CLASS_STRUCT* int_class);
268
270
bool AdaptableWord(WERD_RES* word);
269
271
void EndAdaptiveClassifier();
270
void PrintAdaptiveStatistics(FILE *File);
271
272
void SettupPass1();
272
273
void SettupPass2();
273
274
void AdaptiveClassifier(TBLOB *Blob,
276
277
void ClassifyAsNoise(ADAPT_RESULTS *Results);
277
278
void ResetAdaptiveClassifierInternal();
278
279
279
int GetBaselineFeatures(TBLOB *Blob,
280
INT_TEMPLATES Templates,
281
INT_FEATURE_ARRAY IntFeatures,
282
uinT8* CharNormArray,
283
inT32 *BlobLength);
284
int GetCharNormFeatures(TBLOB *Blob,
285
INT_TEMPLATES Templates,
286
INT_FEATURE_ARRAY IntFeatures,
287
uinT8* PrunerNormArray,
288
uinT8* CharNormArray,
289
inT32 *BlobLength);
280
int GetCharNormFeature(const INT_FX_RESULT_STRUCT& fx_info,
281
INT_TEMPLATES templates,
282
uinT8* pruner_norm_array,
283
uinT8* char_norm_array);
290
284
// Computes the char_norm_array for the unicharset and, if not NULL, the
291
285
// pruner_array as appropriate according to the existence of the shape_table.
292
286
// The norm_feature is deleted as it is almost certainly no longer needed.
298
292
bool TempConfigReliable(CLASS_ID class_id, const TEMP_CONFIG &config);
299
293
void UpdateAmbigsGroup(CLASS_ID class_id, TBLOB *Blob);
300
294
301
void ResetFeaturesHaveBeenExtracted();
302
295
bool AdaptiveClassifierIsFull() { return NumAdaptationsFailed > 0; }
303
296
bool LooksLikeGarbage(TBLOB *blob);
304
297
void RefreshDebugWindow(ScrollView **win, const char *msg,
468
461
469
462
// Create dummy proto and config masks for use with the built-in templates.
470
463
BIT_VECTOR AllProtosOn;
471
BIT_VECTOR PrunedProtos;
472
464
BIT_VECTOR AllConfigsOn;
473
BIT_VECTOR AllProtosOff;
474
465
BIT_VECTOR AllConfigsOff;
475
466
BIT_VECTOR TempProtoMask;
476
467
bool EnableLearning;
504
495
ShapeTable* shape_table_;
505
496
506
497
private:
507
508
498
Dict dict_;
509
499
// The currently active static classifier.
510
500
ShapeClassifier* static_classifier_;
511
501
512
502
/* variables used to hold performance statistics */
513
int AdaptiveMatcherCalls;
514
int BaselineClassifierCalls;
515
int CharNormClassifierCalls;
516
int AmbigClassifierCalls;
517
int NumWordsAdaptedTo;
518
int NumCharsAdaptedTo;
519
int NumBaselineClassesTried;
520
int NumCharNormClassesTried;
521
int NumAmbigClassesTried;
522
int NumClassesOutput;
523
503
int NumAdaptationsFailed;
524
504
525
/* variables used to hold onto extracted features. This is used
526
to map from the old scheme in which baseline features and char norm
527
features are extracted separately, to the new scheme in which they
528
are extracted at the same time. */
529
bool FeaturesHaveBeenExtracted;
530
bool FeaturesOK;
531
INT_FEATURE_ARRAY BaselineFeatures;
532
INT_FEATURE_ARRAY CharNormFeatures;
533
INT_FX_RESULT_STRUCT FXInfo;
534
535
505
// Expected number of features in the class pruner, used to penalize
536
506
// unknowns that have too few features (like a c being classified as e) so
537
507
// it doesn't recognize everything as '@' or '#'.
Loggerhead is a web-based interface for Breezy
Version: 2.0.1