Hi!

Thanks for packaging tesseract!

I was looking for the training scripts and found them in
libtesseract-dev, which doesn't quite make sense to me: they are not
library headers or include files.  (I am referring to the three
scripts in /usr/share/tesseract-ocr, namely language-specific.sh,
tesstrain.sh and tesstrain_utils.sh.)

At the same time, the training binaries are in tesseract-ocr, such as
classifier_tester, lstmtraining and so on.  Would it not make more
sense to have *only* /usr/bin/tesseract in tesseract-ocr, and all of
the other binaries, along with the shell scripts noted above, in a
separate package called something like tesseract-training?  Few people
will be interested in these training scripts, and those who are will
probably want the tesstrain.sh script as well.

Note, though, that the shell scripts tesstrain_utils.sh and
language-specific.sh) are called from tesstrain.sh using $(dirname
$0); that could be fixed by putting tesstrain.sh as /usr/bin/tesstrain
and changing the references to the subscripts to explicitly refer to
/usr/share/tesseract-ocr/.

Best wishes,

   Julian

On Sun, 04 Aug 2019 18:49:24 +0100 Julian Gilbey <jdg@debian.org> wrote:
[...]

Hi Julian,

it makes sense to put the executables which are required for training
into an extra package - not for 4.x because that would break existing
installations, but for 5.x.

The training shell scripts are less relevant and might be replaced
in the future by the already existing Python scripts.

In addition, more and more people run Tesseract training without
that scripts. Instead of those scripts, they use the Makefile
provided by https://github.com/tesseract-ocr/tesstrain/.

Regards,
Stefan