We can probably remove both the old segment-region/line/word and new (all-in-one) segment altogether now that we can configure them via overwrite_* and textequiv_level in recognize. Or we keep the CLI names, but delegate to recognize @kba?

The memory usage of ocrd-tesserocr-segment-region increases for each page, resulting in a total of about 7 GB for 200 pages, 8 GB for 248 pages, 10 GB for 282 pages, 11 GB for 313 pages (observed for http://nbn-resolving.de/urn:nbn:de:bsz:180-digad-22977).

ocrd-tesserocr-segment-line shows a similar effect.

For that book, a machine with 8 GB RAM would have started swapping, thus slowing down the process extremely. Even a large server would get memory problems when processing large books with more than 1000 pages in parallel.

This fixes #101 (using raw_lines by default for textline images, but there are still some corner cases that need to be fixed in Tesseract) and brings a number of segmentation-related improvements:

• interprete overwrite_regions more consistently
• annotate @orientation (independent of dedicated deskewing processor) for vertical and @type for all other text blocks
• no separators and noise regions in reading order
• segment tables into cells and lines so they can be OCRed, too

21:19:10.443 INFO processor.TesserocrSegmentLine - INPUT FILE 65 / phys396119
21:19:10.577 INFO processor.TesserocrSegmentLine - Page 'phys396119' images will use DPI estimated from segmentation
21:19:10.850 ERROR shapely.geos - TopologyException: Input geom 1 is invalid: Self-intersection at or near point 0 107 at 0 107
Traceback (most recent call last):
  File "/home/stweil/src/github/OCR-D/venv-20200904/bin/ocrd-tesserocr-segment-line", line 8, in <module>
    sys.exit(ocrd_tesserocr_segment_line())
  File "/home/stweil/src/github/OCR-D/venv-20200904/lib/python3.7/site-packages/click/core.py", line 829, in __call__
    return self.main(*args, **kwargs)
  File "/home/stweil/src/github/OCR-D/venv-20200904/lib/python3.7/site-packages/click/core.py", line 782, in main
    rv = self.invoke(ctx)
  File "/home/stweil/src/github/OCR-D/venv-20200904/lib/python3.7/site-packages/click/core.py", line 1066, in invoke
    return ctx.invoke(self.callback, **ctx.params)
  File "/home/stweil/src/github/OCR-D/venv-20200904/lib/python3.7/site-packages/click/core.py", line 610, in invoke
    return callback(*args, **kwargs)
  File "/home/stweil/src/github/OCR-D/venv-20200904/lib/python3.7/site-packages/ocrd_tesserocr/cli.py", line 26, in ocrd_tesserocr_segment_line
    return ocrd_cli_wrap_processor(TesserocrSegmentLine, *args, **kwargs)
  File "/home/stweil/src/github/OCR-D/venv-20200904/lib/python3.7/site-packages/ocrd/decorators.py", line 102, in ocrd_cli_wrap_processor
    run_processor(processorClass, ocrd_tool, mets, workspace=workspace, **kwargs)
  File "/home/stweil/src/github/OCR-D/venv-20200904/lib/python3.7/site-packages/ocrd/processor/helpers.py", line 69, in run_processor
    processor.process()
  File "/home/stweil/src/github/OCR-D/venv-20200904/lib/python3.7/site-packages/ocrd_tesserocr/segment_line.py", line 119, in process
    interline = line_poly.intersection(region_poly)
  File "/home/stweil/src/github/OCR-D/venv-20200904/lib/python3.7/site-packages/shapely/geometry/base.py", line 676, in intersection
    return geom_factory(self.impl['intersection'](self, other))
  File "/home/stweil/src/github/OCR-D/venv-20200904/lib/python3.7/site-packages/shapely/topology.py", line 70, in __call__
    self._check_topology(err, this, other)
  File "/home/stweil/src/github/OCR-D/venv-20200904/lib/python3.7/site-packages/shapely/topology.py", line 38, in _check_topology
    self.fn.__name__, repr(geom)))
shapely.errors.TopologicalError: The operation 'GEOSIntersection_r' could not be performed. Likely cause is invalidity of the geometry <shapely.geometry.polygon.Polygon object at 0x7f89253f7c88>

This is just a proof-of-concept that it is possible to load tesseract models installed with ocrd resmgr into the cache directory.

The tricky part here is that there is only one TESSDATA_PREFIX but potentially multiple directories with models. So while it is no problem to look up models in various folders, only one of the can be used as the TESSDATA_PREFIX. Suggestions for a reasonable resolution to this dilemma are welcome.

This is an attempt to implement the other annotation levels. In my opinion, the behaviour for the different levels cannot be made completely analogous with Tesseract: simply pointing it to rectangles for words and glyphs (from an external layout segmentation) would produce results of far worse quality than always recognizing one complete line and allowing its own segmentation below it (accessible by iterators). In contrast, from the line level upwards we can reliably use its respective page segmentation mode (SINGLE_LINE / SINGLE_BLOCK / AUTO). Perhaps warnings and exceptions should be dealt with in a different, more systematic way though.

• all: use second output position as fileGrp USE to produce AlternativeImage
• all: rid of MetadataItem/Labels-related FIXME: with the updated PAGE model, we can now use @externalModel and @externalId
• all: use OcrdExif.resolution instead of xResolution
• all: create images with monotonically growing @comments (features)
• crop: use ocrd_utils.crop_image instead of PIL.Image.crop
• crop: fix bug when trying to access page_image if there are already region coordinates that we are ignoring
• crop: filter images already deskewed and cropped! (we must crop ourselves, and deskewing can not happen until afterwards)
• deskew: fix bugs in configuration-dependent corner cases related to whether deskewing has already been applied (on the page or region level):
  • for the page image, never use images already rotated (both for page level and region level processing, but for the region level, do rotate images ad hoc if @orientation is present on the page level)
  • for the region image, never use images already rotated (except for our own page-level rotation)
• segment-region: forgot to add feature "cropped" when producing cropped images

I use pip install ocrd_tesserocr to install ocrd_tesseract into my virtualenv environment. The installation fails with:

...
  Running setup.py bdist_wheel for tesserocr ... error
  Complete output from command /run/media/flo/a57ed1c0-7fc5-41b1-a6e5-0d43b3ae6a40/data/devel/work/cis-ocrd-py/env/bin/python -u -c "import setuptools, tokenize;__file__='/tmp/pip-install-k_dgo547/tesserocr/setup.py';f=getattr(tokenize, 'open', open)(__file__);code=f.read().replace('\r\n', '\n');f.close();exec(compile(code, __file__, 'exec'))" bdist_wheel -d /tmp/pip-wheel-q7mozwr8 --python-tag cp37:
  Supporting tesseract v4.0.0
  Configs from pkg-config: {'include_dirs': ['/usr/include'], 'libraries': ['tesseract', 'lept'], 'cython_compile_time_env': {'TESSERACT_VERSION': 67108864}}
  running bdist_wheel
  running build
  running build_ext
  building 'tesserocr' extension
  creating build
  creating build/temp.linux-x86_64-3.7
  gcc -pthread -Wno-unused-result -Wsign-compare -DNDEBUG -g -fwrapv -O3 -Wall -march=x86-64 -mtune=generic -O3 -pipe -fstack-protector-strong -fno-plt -march=x86-64 -mtune=generic -O3 -pipe -fstack-protector-strong -fno-plt -march=x86-64 -mtune=generic -O3 -pipe -fstack-protector-strong -fno-plt -fPIC -I/usr/include -I/usr/include/python3.7m -c tesserocr.cpp -o build/temp.linux-x86_64-3.7/tesserocr.o -std=c++11 -DUSE_STD_NAMESPACE
  tesserocr.cpp: In function 'PyObject* __pyx_pf_9tesserocr_16PyResultIterator_8GetBestLSTMSymbolChoices(__pyx_obj_9tesserocr_PyResultIterator*)':
  tesserocr.cpp:12196:43: error: 'class tesseract::ResultIterator' has no member named 'GetBestLSTMSymbolChoices'
     __pyx_v_output = (__pyx_v_self->_riter->GetBestLSTMSymbolChoices()[0]);
                                             ^~~~~~~~~~~~~~~~~~~~~~~~
  error: command 'gcc' failed with exit status 1

  ----------------------------------------
  Failed building wheel for tesserocr
  Running setup.py clean for tesserocr
Failed to build tesserocr
Installing collected packages: tesserocr, ocrd-tesserocr
  Running setup.py install for tesserocr ... error
    Complete output from command /run/media/flo/a57ed1c0-7fc5-41b1-a6e5-0d43b3ae6a40/data/devel/work/cis-ocrd-py/env/bin/python -u -c "import setuptools, tokenize;__file__='/tmp/pip-install-k_dgo547/tesserocr/setup.py';f=getattr(tokenize, 'open', open)(__file__);code=f.read().replace('\r\n', '\n');f.close();exec(compile(code, __file__, 'exec'))" install --record /tmp/pip-record-fc87h61b/install-record.txt --single-version-externally-managed --compile --install-headers /run/media/flo/a57ed1c0-7fc5-41b1-a6e5-0d43b3ae6a40/data/devel/work/cis-ocrd-py/env/include/site/python3.7/tesserocr:
    Supporting tesseract v4.0.0
    Configs from pkg-config: {'include_dirs': ['/usr/include'], 'libraries': ['lept', 'tesseract'], 'cython_compile_time_env': {'TESSERACT_VERSION': 67108864}}
    running install
    running build
    running build_ext
    building 'tesserocr' extension
    creating build
    creating build/temp.linux-x86_64-3.7
    gcc -pthread -Wno-unused-result -Wsign-compare -DNDEBUG -g -fwrapv -O3 -Wall -march=x86-64 -mtune=generic -O3 -pipe -fstack-protector-strong -fno-plt -march=x86-64 -mtune=generic -O3 -pipe -fstack-protector-strong -fno-plt -march=x86-64 -mtune=generic -O3 -pipe -fstack-protector-strong -fno-plt -fPIC -I/usr/include -I/usr/include/python3.7m -c tesserocr.cpp -o build/temp.linux-x86_64-3.7/tesserocr.o -std=c++11 -DUSE_STD_NAMESPACE
    tesserocr.cpp: In function 'PyObject* __pyx_pf_9tesserocr_16PyResultIterator_8GetBestLSTMSymbolChoices(__pyx_obj_9tesserocr_PyResultIterator*)':
    tesserocr.cpp:12196:43: error: 'class tesseract::ResultIterator' has no member named 'GetBestLSTMSymbolChoices'
       __pyx_v_output = (__pyx_v_self->_riter->GetBestLSTMSymbolChoices()[0]);
                                               ^~~~~~~~~~~~~~~~~~~~~~~~
    error: command 'gcc' failed with exit status 1
...

tesseract is installed on the system:

tesseract 4.0.0-beta.4-26-gfd49
 leptonica-1.77.0
  libgif 5.1.4 : libjpeg 8d (libjpeg-turbo 2.0.1) : libpng 1.6.36 : libtiff 4.0.10 : zlib 1.2.11 : libwebp 1.0.1
 Found AVX
 Found SSE

At the moment, superfluous newlines are appended to the TextEquiv/Unicode entries:

                    <pc:TextEquiv>
                        <pc:Unicode>Groſzmaͤchtigſter</pc:Unicode>
                    </pc:TextEquiv>
                    <pc:TextEquiv>
                        <pc:Unicode>stzmächtigstcr
</pc:Unicode>

Since Tesseract 4, two OCR engines are available: rule-based (i.e. --oem 0), LSTM (--oem 1). The command-line also exposes an ensemble of the two OCR engines (--oem 2). The documentation for ocrd-tesserocr-recognize does not make it clear which engine is used and using either the following parameters seems to have no effect on the recognition results:

• -P tesseract_parameters '{ "tessedit_ocr_engine_mode" : "0" }'
• -P tesseract_parameters '{ "tessedit_ocr_engine_mode" : "1" }'
• -P tesseract_parameters '{ "tessedit_ocr_engine_mode" : "2" }'

Which one of the OCR engines are we currently using?

And with this image:

https://digi.ub.uni-heidelberg.de/diglitData/v/justinian1627bd2_-_1281.tif

and ocrd.sif (singularity container) created from docker ocrd_all at Nov 9 10:13 2021 & at Jan 17 15:11 2022 [UPDATE]

and this workflow:

/usr/bin/time singularity exec $HOME/ocrd.sif ocrd workspace init >>ocrd.log 2>&1 || exit
/usr/bin/time singularity exec $HOME/ocrd.sif ocrd workspace add -g P_00001 -G OCR-D-IMG -i OCR-D-IMG_00001 -m image/tiff OCR-D-IMG/00001.tif >>ocrd.log 2>&1 || exit

/usr/bin/time singularity exec $HOME/ocrd.sif ocrd-olena-binarize -P k 0.10 -I OCR-D-IMG -O OCR-D-001 >>ocrd.log 2>&1 || exit
/usr/bin/time singularity exec $HOME/ocrd.sif ocrd-anybaseocr-crop -I OCR-D-001 -O OCR-D-002 >>ocrd.log 2>&1 || exit
/usr/bin/time singularity exec $HOME/ocrd.sif ocrd-olena-binarize -I OCR-D-002 -O OCR-D-003 >>ocrd.log 2>&1 || exit
/usr/bin/time singularity exec $HOME/ocrd.sif ocrd-cis-ocropy-deskew -P level-of-operation page -I OCR-D-003 -O OCR-D-004 >>ocrd.log 2>&1 || exit
/usr/bin/time singularity exec $HOME/ocrd.sif ocrd-tesserocr-segment -P find_tables false -P shrink_polygons true -I OCR-D-004 -O OCR-D-005 >>ocrd.log 2>&1 || exit
/usr/bin/time singularity exec $HOME/ocrd.sif ocrd-calamari-recognize -I OCR-D-005 -O OCR-D-OCR -P checkpoint "$HOME/ocrd_models/calamari/calamari_models_experimental/historical_french_2020-10-14/*.ckpt.json" >>ocrd.log 2>&1 || exit

I'll get a segmentation fault

Core was generated by `/usr/bin/python3 /usr/bin/ocrd-tesserocr-segment -P find_tables false -P shrink'.
Program terminated with signal 11, Segmentation fault.

when using for example Arabic model, recognition works fine but the words inside the generated PAGE-XML contains reversed letters. But the sequence of words itself is correct, here an example: generated word with wrong sequence of letters:

               <pc:Word id="region0001_line0001_word0000">
                    <pc:Coords points="1620,372 1620,402 1703,402 1703,375 1647,376"/>
                    <pc:TextEquiv conf="0.877831573486328">
                        <pc:Unicode>رصم</pc:Unicode>
                    </pc:TextEquiv>
                </pc:Word>

but the line containing the recogized word should look like this:

                        <pc:Unicode>مصر</pc:Unicode>

(I know it is not easy to see clearly that it is reversed because the letters in Arabic changes appearance depending on position inside word, but this is handled by font.)

Here is the equivalent portion of the image:

REMARK: when using tesseract as standalone and generating alto, the sequence is correct!

this image

https://digi.ub.uni-heidelberg.de/diglitData/v/montfaucon1719bd2_1.210.tif

UPDATE same for https://digi.ub.uni-heidelberg.de/diglitData/v/montfaucon1719bd2_1.168a_Planche_72.tif

with this workflow (latest ocrd_all as of 2021-12-01)

ocrd workspace init 
ocrd workspace add -g P_00001 -G OCR-D-IMG -i OCR-D-IMG_00001 -m image/tiff OCR-D-IMG/00001.tif 

ocrd-olena-binarize -P k 0.10 -I OCR-D-IMG -O OCR-D-001 
ocrd-anybaseocr-crop -I OCR-D-001 -O OCR-D-002 
ocrd-olena-binarize -I OCR-D-002 -O OCR-D-003 
ocrd-cis-ocropy-deskew -P level-of-operation page -I OCR-D-003 -O OCR-D-004 
ocrd-tesserocr-segment -P find_tables false -P shrink_polygons true -I OCR-D-004 -O OCR-D-005 
ocrd-calamari-recognize -I OCR-D-005 -O OCR-D-OCR -P checkpoint "$HOME/ocrd/_models/ocrd-calamari-recognize/c1_latin-script-hist-3/*.ckpt.json" 

leads to this error messages:

10:06:58.121 INFO processor.TesserocrSegment - INPUT FILE 0 / P_00001
10:06:59.193 INFO processor.TesserocrSegment - Page 'P_00001' images will use 333 DPI from image 
meta-data
10:06:59.193 INFO processor.TesserocrSegment - Processing page 'P_00001'
10:07:00.229 INFO ocrd.workspace.save_image_file - created file ID: OCR-D-005_00001.IMG-BIN, 
file_grp: OCR-D-005, path: OCR-D-005/OCR-D-005_00001.IMG-BIN.png
/build/ocrd_tesserocr/ocrd_tesserocr/recognize.py:510: ShapelyDeprecationWarning: The proxy 
geometries (through the 'asShape()', 'asPolygon()' or 'PolygonAdapter()' constructors) are 
deprecated and will be removed in Shapely 2.0. Use the 'shape()' function or the standard 
'Polygon()' constructor instead.
  for symbol in iterate_level(it, RIL.SYMBOL, parent=RIL.BLOCK)])
Exception ignored in: <bound method BaseGeometry.__del__ of 
<shapely.geometry.polygon.PolygonAdapter object at 0x7fc431060358>>
Traceback (most recent call last):
  File "/usr/local/lib/python3.6/dist-packages/shapely/geometry/base.py", line 209, in __del__
    self._empty(val=None)
  File "/usr/local/lib/python3.6/dist-packages/shapely/geometry/base.py", line 199, in _empty
    self._is_empty = True
  File "/usr/local/lib/python3.6/dist-packages/shapely/geometry/proxy.py", line 44, in __setattr__
    object.__setattr__(self, name, value)
AttributeError: can't set attribute
10:07:00.930 INFO processor.TesserocrSegment - Detected region 'region0000': 2867,801 2418,798 
1883,799 1527,803 1527,803 1184,824 1184,824 1183,824 1183,824 1183,824 1183,824 1183,824 1183,825 
1181,827 1180,827 1180,827 1180,827 1180,827 1180,827 1180,828 1180,828 1180,828 1180,838 1172,2362 
1171,3063 1175,3451 1175,3451 1175,3451 1175,3452 1175,3452 1175,3452 1175,3452 1175,3452 1176,3452 
1176,3453 1176,3453 1176,3453 1176,3453 1176,3453 1177,3453 1260,3474 1260,3474 1260,3474 1304,3474 
1945,3458 1945,3458 3324,3389 3324,3389 3325,3389 3348,3382 3348,3382 3348,3382 3348,3382 3348,3382 
3348,3381 3349,3381 3349,3381 3349,3381 3349,3381 3349,3381 3349,3380 3349,3380 3349,3380 3387,1134 
3388,1069 3388,1069 3377,954 3377,954 3377,953 3377,953 3377,953 3377,953 3354,913 3354,913 
3353,913 3353,912 3353,912 3353,912 3353,912 3130,804 3130,804 3129,804 3129,804 3129,804 
(FLOWING_TEXT)
...
...
...
Exception ignored in: <bound method BaseGeometry.__del__ of 
<shapely.geometry.polygon.PolygonAdapter object at 0x7fc40f820710>>
Traceback (most recent call last):
  File "/usr/local/lib/python3.6/dist-packages/shapely/geometry/base.py", line 209, in __del__
    self._empty(val=None)
  File "/usr/local/lib/python3.6/dist-packages/shapely/geometry/base.py", line 199, in _empty
    self._is_empty = True
  File "/usr/local/lib/python3.6/dist-packages/shapely/geometry/proxy.py", line 44, in __setattr__
    object.__setattr__(self, name, value)
AttributeError: can't set attribute
10:07:16.823 INFO processor.TesserocrSegment - Detected line 'region0005_line0010': 2366,4729 
2366,4729 2366,4729 2291,4740 2290,4740 2290,4740 2290,4740 2290,4740 2290,4740 2290,4741 2289,4741 
2289,4741 2289,4741 2289,4741 2289,4741 2289,4742 2289,4742 2289,4742 2289,4780 2289,4780 2289,4780 
2289,4781 2289,4781 2289,4781 2289,4781 2289,4781 2290,4781 2290,4782 2290,4782 2290,4782 2290,4782 
2290,4782 2291,4782 2291,4782 2291,4782 2650,4795 2895,4801 2905,4801 2905,4801 3188,4781 3188,4781 
3189,4781 3189,4781 3189,4781 3189,4781 3189,4781 3189,4780 3190,4780 3190,4780 3190,4780 3190,4780 
3190,4780 3190,4779 3190,4779 3190,4779 3190,4768 3190,4768 3190,4768 3190,4767 3190,4767 3190,4767 
3190,4767 3190,4767 3189,4767 3189,4766 3189,4766 3189,4766 3189,4766 3189,4766 3188,4766 3188,4766 
2705,4736 2705,4736 2638,4732
Traceback (most recent call last):
  File "/usr/local/sub-venv/headless-tf2/bin/ocrd-calamari-recognize", line 33, in <module>
    sys.exit(load_entry_point('ocrd-calamari', 'console_scripts', 'ocrd-calamari-recognize')())
  File "/usr/local/sub-venv/headless-tf2/lib/python3.6/site-packages/click/core.py", line 1128, in 
__call__
    return self.main(*args, **kwargs)
  File "/usr/local/sub-venv/headless-tf2/lib/python3.6/site-packages/click/core.py", line 1053, in 
main
    rv = self.invoke(ctx)
  File "/usr/local/sub-venv/headless-tf2/lib/python3.6/site-packages/click/core.py", line 1395, in 
invoke
    return ctx.invoke(self.callback, **ctx.params)
  File "/usr/local/sub-venv/headless-tf2/lib/python3.6/site-packages/click/core.py", line 754, in 
invoke
    return __callback(*args, **kwargs)
  File "/build/ocrd_calamari/ocrd_calamari/cli.py", line 13, in ocrd_calamari_recognize
    return ocrd_cli_wrap_processor(CalamariRecognize, *args, **kwargs)
  File "/build/core/ocrd/ocrd/decorators/__init__.py", line 90, in ocrd_cli_wrap_processor
    raise Exception("Invalid input/output file grps:\n\t%s" % '\n\t'.join(report.errors))
Exception: Invalid input/output file grps:
        Input fileGrp[@USE='OCR-D-005'] not in METS!
```

The current code imports numpy although it only uses a single function from that library. Including numpy creates a number of threads for the BLAS algorithms by default. Those threads use a lot of CPU time without doing anything useful.

Setting the environment variable OMP_THREAD_LIMIT=1 avoids those additional threads.

Maybe there exists a better solution which does not require an environment variable, for example removing the numpy requirement.

With tables where there are no horizontal lines, the workflow results in a wrong reading order by only recognizing the columns and no rows.
See the following image as an example:

The result is as follows: OCR-D-TXT_catalog46muse_0564.txt

This is the used workfow:

ocrd-olena-binarize -I OCR-D-OPT -O OCR-D-BIN -p '{"impl": "sauvola-ms-split"}'
ocrd-cis-ocropy-denoise -I OCR-D-BIN -O OCR-D-DENOISE -p '{"level-of-operation":"page"}'
ocrd-cis-ocropy-deskew -I OCR-D-DENOISE -O OCR-D-DESKEW-PAGE -p '{"level-of-operation":"page"}'
ocrd-tesserocr-segment-region -I OCR-D-DESKEW-PAGE -O OCR-D-SEG-REG
ocrd-segment-repair -I OCR-D-SEG-REG -O OCR-D-SEG-REPAIR -p '{"plausibilize":true}'
ocrd-cis-ocropy-binarize -I OCR-D-SEG-REPAIR -O OCR-D-BIN2 -p '{"level-of-operation":"region"}'
ocrd-tesserocr-deskew -I OCR-D-BIN2 -O OCR-D-DESKEW-TEXT
ocrd-tesserocr-segment-line -I OCR-D-DESKEW-TEXT -O OCR-D-SEG-LINE
ocrd-cis-ocropy-resegment -I OCR-D-SEG-LINE -O OCR-D-RESEG
ocrd-cis-ocropy-dewarp -I OCR-D-RESEG -O OCR-D-DEWARP-LINE
ocrd-tesserocr-recognize -I OCR-D-DEWARP-LINE -O OCR-D-OCR -p '{"model": "deu"}'