Add option --naming for different line name patterns

[zuphilip]

Usually the lines are named as a hexadecimal 6-digits number where the first two digits are always 01 starting from 010001.bin.png. However, the hexadecimal numbers are sometimes sorting differently and don't fulfill an easy regexp as ´\d+`.

This new option for ocropus-gpageseg will introduce the possibility to change the naming patterns. By default args.naming="hex", we have the same behavior as at the moment. Moreover, there is the option to change to a decimal naming pattern with --naming dec, which will produce 4-digits (decimal) numbers starting from 0001.bin.png. Any other value will be tried to be interpreted as an individual pattern, e.g. --naming ex%05d.

Note: The pseg format will save the page segmentation by labeling the different lines with 6-digits hexadecimal numbers which are then represented as a number. Therefore, ocropus-hocr needed also a change. Are there more consequences when not using 6-digits hexadecimal numbers for naming the lines?

Any other reaction for this idea?

[wrznr]

This a very valuable enhancement. However, I would drop the predefined choices and rely on pure format strings only.

[zuphilip]

@wrznr Do you mean that with the naming option only a user-defined pattern can be indicated (deleting the hex and dec cases)? Then, this parameter should be optional and by default the hexadecimal pattern be used as it is currently (to not break any possible dependencies on tools based on ocropus).

Then, these calls would be possible:

ocropus-gpageseg test.bin.png
# by default the pattern "01%04x" is used

ocropus-gpageseg test.bin.png --naming %04d
# pattern with 4 decimal digits

[zuphilip]

There may be more places to attribute different length of file names:

root@zuphilip-VirtualBox:/etc/ocropy# grep -r \?\?\?\?\?\? *
ocrolib/common.py:        return sorted(glob.glob(args[0]+"/????/??????.png"))
ocropus-hocr:xhfiles = python.sum([glob.glob(d+"/??????.xheight") for d in dirs],[])
ocropus-hocr:    lfiles = python.sum([glob.glob(d+"/??????.bin.png") for d in dirs],[])
ocropus-visualize-results:            images = sorted(glob.glob("??????.bin.png"))
ocropus-visualize-results:        for fname in sorted(glob.glob(d+"/??????.txt")):
README.md:    ./ocropus-rpred -Q 4 -m models/fraktur.pyrnn.gz 'book/????/??????.bin.png'
run-coverage:python -m coverage run -p ocropus-rpred -n 'temp/????/??????.bin.png'
run-coverage:python -m coverage run -p ocropus-gtedit html temp/????/??????.bin.png -o temp-correction.html
run-test:ocropus-rpred -n 'temp/????/??????.bin.png'
run-test:ocropus-gtedit html temp/????/??????.bin.png -o temp-correction.html
run-test-ci:ocropus-gtedit html temp/????/??????.bin.png -o temp-correction.html

[amitdo]

https://www.mail-archive.com/[email protected]/msg00807.html
https://www.mail-archive.com/[email protected]/msg00850.html

[zuphilip]

@amitdo What do you want to point me to with these two links?

[amitdo]

At some point in history, Tom switched from %04d to %04x.

Just digging in ocropus history...