Method SetupForClustering

deps/tesseract/classify/mastertrainer.cpp:550–584

Sets up a Clusterer for mftraining on a single shape_id. Call FreeClusterer on the return value after use.

// Sets up a Clusterer for mftraining on a single shape_id.
// Call FreeClusterer on the return value after use.
CLUSTERER* MasterTrainer::SetupForClustering(
    const ShapeTable& shape_table,
    const FEATURE_DEFS_STRUCT& feature_defs,
    int shape_id,
    int* num_samples) {

  int desc_index = ShortNameToFeatureType(feature_defs, kMicroFeatureType);
  int num_params = feature_defs.FeatureDesc[desc_index]->NumParams;
  ASSERT_HOST(num_params == MFCount);
  CLUSTERER* clusterer = MakeClusterer(
      num_params, feature_defs.FeatureDesc[desc_index]->ParamDesc);

  // We want to iterate over the samples of just the one shape.
  IndexMapBiDi shape_map;
  shape_map.Init(shape_table.NumShapes(), false);
  shape_map.SetMap(shape_id, true);
  shape_map.Setup();
  // Reverse the order of the samples to match the previous behavior.
  GenericVector<const TrainingSample*> sample_ptrs;
  SampleIterator it;
  it.Init(&shape_map, &shape_table, false, &samples_);
  for (it.Begin(); !it.AtEnd(); it.Next()) {
    sample_ptrs.push_back(&it.GetSample());
  }
  int sample_id = 0;
  for (int i = sample_ptrs.size() - 1; i >= 0; --i) {
    const TrainingSample* sample = sample_ptrs[i];
    int num_features = sample->num_micro_features();
    for (int f = 0; f < num_features; ++f)
      MakeSample(clusterer, sample->micro_features()[f], sample_id);
    ++sample_id;
  }
  *num_samples = sample_id;
  return clusterer;
}

// Writes the given float_classes (produced by SetupForFloat2Int) as inttemp
// to the given inttemp_file, and the corresponding pffmtable.