MCPcopy Create free account
hub / github.com/creatale/node-dv / ComputeCanonicalFeatures

Method ComputeCanonicalFeatures

deps/tesseract/classify/trainingsampleset.cpp:791–805  ·  view source on GitHub ↗

Caches the indexed features of the canonical samples. ComputeCanonicalSamples must have been already called. TODO(rays) see note on ReliablySeparable and try restricting the canonical features to those that truly represent all samples.

Source from the content-addressed store, hash-verified

789// TODO(rays) see note on ReliablySeparable and try restricting the
790// canonical features to those that truly represent all samples.
791void TrainingSampleSet::ComputeCanonicalFeatures() {
792 ASSERT_HOST(font_class_array_ != NULL);
793 int font_size = font_id_map_.CompactSize();
794 for (int font_index = 0; font_index < font_size; ++font_index) {
795 int font_id = font_id_map_.CompactToSparse(font_index);
796 for (int c = 0; c < unicharset_size_; ++c) {
797 int num_samples = NumClassSamples(font_id, c, false);
798 if (num_samples == 0)
799 continue;
800 const TrainingSample* sample = GetCanonicalSample(font_id, c);
801 FontClassInfo& fcinfo = (*font_class_array_)(font_index, c);
802 fcinfo.canonical_features = sample->indexed_features();
803 }
804 }
805}
806
807// Computes the combined set of features used by all the samples of each
808// font/class combination. Use after ReplicateAndRandomizeSamples.

Callers 1

PreTrainingSetupMethod · 0.80

Calls 2

CompactSizeMethod · 0.80
CompactToSparseMethod · 0.80

Tested by

no test coverage detected