// Constructor is private. Only anticipated use of ErrorCounter is via
// the static ComputeErrorRate.
ErrorCounter::ErrorCounter(const UNICHARSET& unicharset, int fontsize)
: scaled_error_(0.0), rating_epsilon_(kRatingEpsilon),
unichar_counts_(unicharset.size(), unicharset.size(), 0),
ok_score_hist_(0, 101), bad_score_hist_(0, 101),
unicharset_(unicharset) {
Counts empty_counts;
font_counts_.init_to_size(fontsize, empty_counts);
multi_unichar_counts_.init_to_size(unicharset.size(), 0);
}
// Helper to set the properties for an input unicharset file, writes to the
// output file. If an appropriate script unicharset can be found in the
// script_dir directory, then the tops and bottoms are expanded using the
// script unicharset.
// If non-empty, xheight data for the fonts are written to the xheights_file.
void SetPropertiesForInputFile(const std::string& script_dir,
const std::string& input_unicharset_file,
const std::string& output_unicharset_file,
const std::string& output_xheights_file) {
UNICHARSET unicharset;
// Load the input unicharset
unicharset.load_from_file(input_unicharset_file.c_str());
tprintf("Loaded unicharset of size %d from file %s\n", unicharset.size(),
input_unicharset_file.c_str());
// Set unichar properties
tprintf("Setting unichar properties\n");
SetupBasicProperties(true, false, &unicharset);
tprintf("Setting script properties\n");
SetScriptProperties(script_dir, &unicharset);
if (!output_xheights_file.empty()) {
std::string xheights_str = GetXheightString(script_dir, unicharset);
File::WriteStringToFileOrDie(xheights_str, output_xheights_file);
}
// Write the output unicharset
tprintf("Writing unicharset to file %s\n", output_unicharset_file.c_str());
unicharset.save_to_file(output_unicharset_file.c_str());
}
// Helper to set the properties for an input unicharset file, writes to the
// output file. If an appropriate script unicharset can be found in the
// script_dir directory, then the tops and bottoms are expanded using the
// script unicharset.
// If non-empty, xheight data for the fonts are written to the xheights_file.
void SetPropertiesForInputFile(const string& script_dir,
const string& input_unicharset_file,
const string& output_unicharset_file,
const string& output_xheights_file) {
UNICHARSET unicharset;
// Load the input unicharset
unicharset.load_from_file(input_unicharset_file.c_str());
tprintf("Loaded unicharset of size %d from file %s\n", unicharset.size(),
input_unicharset_file.c_str());
// Set unichar properties
tprintf("Setting unichar properties\n");
SetupBasicProperties(true, false, &unicharset);
string xheights_str;
for (int s = 0; s < unicharset.get_script_table_size(); ++s) {
// Load the unicharset for the script if available.
string filename = script_dir + "/" +
unicharset.get_script_from_script_id(s) + ".unicharset";
UNICHARSET script_set;
if (script_set.load_from_file(filename.c_str())) {
unicharset.SetPropertiesFromOther(script_set);
}
// Load the xheights for the script if available.
filename = script_dir + "/" + unicharset.get_script_from_script_id(s) +
".xheights";
string script_heights;
if (File::ReadFileToString(filename, &script_heights))
xheights_str += script_heights;
}
if (!output_xheights_file.empty())
File::WriteStringToFileOrDie(xheights_str, output_xheights_file);
for (int c = SPECIAL_UNICHAR_CODES_COUNT; c < unicharset.size(); ++c) {
if (unicharset.PropertiesIncomplete(c)) {
tprintf("Warning: properties incomplete for index %d = %s\n",
c, unicharset.id_to_unichar(c));
}
}
// Write the output unicharset
tprintf("Writing unicharset to file %s\n", output_unicharset_file.c_str());
unicharset.save_to_file(output_unicharset_file.c_str());
}
int main(int argc, char** argv) {
// Sets properties on the input unicharset file, and writes:
// rootdir/lang/lang.charset_size=ddd.txt
// rootdir/lang/lang.traineddata
// rootdir/lang/lang.unicharset
// If the 3 word lists are provided, the dawgs are also added
// to the traineddata file.
// The output unicharset and charset_size files are just for
// human readability.
tesseract::CheckSharedLibraryVersion();
tesseract::ParseCommandLineFlags(argv[0], &argc, &argv, true);
GenericVector<STRING> words, puncs, numbers;
// If these reads fail, we get a warning message and an empty list of words.
tesseract::ReadFile(FLAGS_words.c_str(), nullptr).split('\n', &words);
tesseract::ReadFile(FLAGS_puncs.c_str(), nullptr).split('\n', &puncs);
tesseract::ReadFile(FLAGS_numbers.c_str(), nullptr).split('\n', &numbers);
// Load the input unicharset
UNICHARSET unicharset;
if (!unicharset.load_from_file(FLAGS_input_unicharset.c_str(), false)) {
tprintf("Failed to load unicharset from %s\n",
FLAGS_input_unicharset.c_str());
return 1;
}
tprintf("Loaded unicharset of size %d from file %s\n", unicharset.size(),
FLAGS_input_unicharset.c_str());
// Set unichar properties
tprintf("Setting unichar properties\n");
tesseract::SetupBasicProperties(/*report_errors*/ true,
/*decompose (NFD)*/ false, &unicharset);
tprintf("Setting script properties\n");
tesseract::SetScriptProperties(FLAGS_script_dir.c_str(), &unicharset);
// Combine everything into a traineddata file.
return tesseract::CombineLangModel(
unicharset, FLAGS_script_dir.c_str(), FLAGS_version_str.c_str(),
FLAGS_output_dir.c_str(), FLAGS_lang.c_str(), FLAGS_pass_through_recoder,
words, puncs, numbers, FLAGS_lang_is_rtl, /*reader*/ nullptr,
/*writer*/ nullptr);
}
