// Classifies the given [training] sample, writing to results.
// See ShapeClassifier for a full description.
int CubeClassifier::ClassifySample(const TrainingSample& sample,
Pix* page_pix, int debug, int keep_this,
GenericVector<ShapeRating>* results) {
results->clear();
if (page_pix == NULL) return 0;
ASSERT_HOST(cube_cntxt_ != NULL);
const TBOX& char_box = sample.bounding_box();
CubeObject* cube_obj = new tesseract::CubeObject(
cube_cntxt_, page_pix, char_box.left(),
pixGetHeight(page_pix) - char_box.top(),
char_box.width(), char_box.height());
CharAltList* alt_list = cube_obj->RecognizeChar();
alt_list->Sort();
CharSet* char_set = cube_cntxt_->CharacterSet();
if (alt_list != NULL) {
for (int i = 0; i < alt_list->AltCount(); ++i) {
// Convert cube representation to a shape_id.
int alt_id = alt_list->Alt(i);
int unichar_id = char_set->UnicharID(char_set->ClassString(alt_id));
int shape_id = shape_table_.FindShape(unichar_id, -1);
if (shape_id >= 0)
results->push_back(ShapeRating(shape_id, alt_list->AltProb(i)));
}
delete alt_list;
}
delete cube_obj;
return results->size();
}
// Classifies the given [training] sample, writing to results.
// See ShapeClassifier for a full description.
int CubeTessClassifier::ClassifySample(const TrainingSample& sample,
Pix* page_pix, int debug, int keep_this,
GenericVector<ShapeRating>* results) {
int num_results = pruner_->ClassifySample(sample, page_pix, debug, keep_this,
results);
if (page_pix == NULL) return num_results;
ASSERT_HOST(cube_cntxt_ != NULL);
const TBOX& char_box = sample.bounding_box();
CubeObject* cube_obj = new tesseract::CubeObject(
cube_cntxt_, page_pix, char_box.left(),
pixGetHeight(page_pix) - char_box.top(),
char_box.width(), char_box.height());
CharAltList* alt_list = cube_obj->RecognizeChar();
CharSet* char_set = cube_cntxt_->CharacterSet();
if (alt_list != NULL) {
for (int r = 0; r < num_results; ++r) {
const Shape& shape = shape_table_.GetShape((*results)[r].shape_id);
// Get the best cube probability of all unichars in the shape.
double best_prob = 0.0;
for (int i = 0; i < alt_list->AltCount(); ++i) {
int alt_id = alt_list->Alt(i);
int unichar_id = char_set->UnicharID(char_set->ClassString(alt_id));
if (shape.ContainsUnichar(unichar_id) &&
alt_list->AltProb(i) > best_prob) {
best_prob = alt_list->AltProb(i);
}
}
(*results)[r].rating = best_prob;
}
delete alt_list;
// Re-sort by rating.
results->sort(&ShapeRating::SortDescendingRating);
}
delete cube_obj;
return results->size();
}
// Returns a string debug representation of the given sample:
// font, unichar_str, bounding box, page.
STRING TrainingSampleSet::SampleToString(const TrainingSample &sample) const {
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;
}
