Closed jwilk closed 9 years ago
For the reference, this is the code that emits this message:
#!c++
if (!sub_langs_.empty()) {
// In multilingual mode word ratings have to be directly comparable,
// so use the same language model weights for all languages:
// use the primary language's params model if
// tessedit_use_primary_params_model is set,
// otherwise use default language model weights.
if (tessedit_use_primary_params_model) {
for (int s = 0; s < sub_langs_.size(); ++s) {
sub_langs_[s]->language_model_->getParamsModel().Copy(
this->language_model_->getParamsModel());
}
tprintf("Using params model of the primary language\n");
if (tessdata_manager_debug_level) {
this->language_model_->getParamsModel().Print();
}
} else {
this->language_model_->getParamsModel().Clear();
for (int s = 0; s < sub_langs_.size(); ++s) {
sub_langs_[s]->language_model_->getParamsModel().Clear();
}
tprintf("Using default language params\n");
}
}
(copied from https://sources.debian.net/src/tesseract/3.03.03-1/ccmain/tessedit.cpp?hl=358#L338)
I'm not sure what the message exactly means.
Comment submitted by @jsbien:
I can ask on the tesseract list.
Comment submitted by @jsbien:
https://groups.google.com/forum/#!topic/tesseract-ocr/Fmz9cPPqb6k
I'm not sure there's much ocrodjvu can do here. I'm leaning towards closing this bug without any action.
Comment submitted by @jsbien:
I understand there is no easy way to mark this message as coming from tesseract, so perhaps just mention it in the documentation?
Actually I could easily add the tesseract:
prefix to everything Tesseract prints on stderr.
Would this make you satisfied?
Comment submitted by @jsbien:
Yes.
Implemented in 71daffaa2fab018d327fa9aaefbf9a1a4149a9f9.
Fixed in ocrodjvu 0.8.
Issue reported by @jsbien:
ocrodjvu -e tesseract -l deu-frak+pol --save-raw-ocr=hocr -o 0184ocr.djvu iLinde1FR11p0184.djvu Processing 'iLinde1FR11p0184.djvu':
...
Tesseract Open Source OCR Engine v3.03 with Leptonica
Using default language params
The message suggests that the language parameters deu-frak and pol has been ignored, but hOCR contains the lang tags applied more or less correctly.
The file used for the test is saved from
http://eprints.wbl.klf.uw.edu.pl/44/1/iLinde1FR11.djvu?djvuopts=&page=184&zoom=width&showposition=0.5,0.2
This is ocrodjvu 0.7.17 on Debian jessie.
Regards
JSB