Request for Pre-trained Handwritten OCR Kraken Models for French or Latin-based Languages

Hello,

I am currently working on an OCR project involving handwritten French text (or other Latin-based European languages would be fine, I guess) and I’m looking for a suitable pre-trained model for use with Kraken. Specifically, I would like a model that performs well with handwritten text in French or similar languages.

Does Kraken have any pre-trained models for French handwriting or Latin-based handwriting OCR? I used chatGPT and MistralIA to find some fr.mlmodel or sth, but all links were broken, I found nothing through google, or in this website...

If not, are there any recommendations for models trained on handwriting datasets (e.g., IAM dataset, Rimes dataset, etc.) that could work well with Kraken?

Any links to shared models, or guidance on how to adapt existing models for this task, would be greatly appreciated.

Thank you for your time and help!

I have currently kraken --version
kraken, version 6.0.2

[PonteIneptique]

Hi !
You can find all models on https://zenodo.org/communities/ocr_models/records?q=&l=list&p=1&s=10&sort=newest , but you can also find them with the kraken command-line
This one is probably your best bet https://zenodo.org/records/14399779

perfect,

downloaded.

I've placed, (gpt told me to place here -- whereas my kraken/java is in C:\Users<name>\AppData\Local\Programs\Python\Python310\Lib\site-packages\kraken) :

C:\Users<name>.local\share\kraken\models\FoNDUE-GD_v2_fr.mlmodel

placed my pngs (1 to 154) in :
C:\Users<name>\Manuscrit_pages\page_1.png , page_2.png, ... page_154.png

tried several commands with segment, baseline and other stuff given to me by chatGPT and MistralIA, but nothing seems to work ...

any good command line at this point that I could use?

[PonteIneptique]

I think reading the documentation yourself might be a good first start, before asking GPT, Mistral or Gemini :) https://kraken.re

thx,

I passed through that one when I was looking for a mlmodel, good point... I started that one again, but still not :

C:\Users<name>\Manuscrit_pages>kraken -i page_1.png page_1.txt segment -bl ocr -m C:/Users//.local/share/kraken/models/FoNDUE-GD_v2_fr.mlmodel
Loading ANN C:\Users<name>\AppData\Local\Programs\Python\Python310\lib\site-packages\kraken\blla.mlmodel ✓
Loading ANN \Users<name>.local\share\kraken\models\FoNDUE-GD_v2_fr.mlmodel ✓
Segmenting page_1.png [12/05/25 13:34:48] WARNING Polygonizer failed on line 0: TopologyException: side location conflict segmentation.py:777
at 1101.6261767385363 272.87853021560886. This can occur if the input
geometry is invalid.
✓
[12/05/25 13:34:50] ERROR Failed processing page_1.png: 'default' kraken.py:448

then I thought, perhaps it's because there is not enough text on page 1 because it's just mostly the title, so I tried page 2,

C:\Users<name>\Manuscrit_pages>kraken -i page_2.png page_2.txt segment -bl ocr -m C:/Users//.local/share/kraken/models/FoNDUE-GD_v2_fr.mlmodel
Loading ANN C:\Users<name>\AppData\Local\Programs\Python\Python310\lib\site-packages\kraken\blla.mlmodel ✓
Loading ANN \Users<name>.local\share\kraken\models\FoNDUE-GD_v2_fr.mlmodel ✓
Segmenting page_2.png ✓
[12/05/25 13:36:07] ERROR Failed processing page_2.png: 'default' kraken.py:448

it doesn't give exactly the same error, but comes from the same line...?

[PonteIneptique]

There is an option at the main command level --raise-on-error or something, that will give you more details on the error

[mittagessen]

The problem is that you're using Windows. The `-m` option can point to a model or a mapping between a line type and model file. The latter separates them with `:` so the Window device is interpreted as a line type. You can try escaping it with `-m C\:/Users/...` and that might work. Kraken does not support Windows though so you might run into other issues.

…

On 25/12/05 04:39AM, ytreza-azerty wrote: ytreza-azerty left a comment (mittagessen/kraken#749) thx, I passed through that one when I was looking for a mlmodel, good point... I started that one again, but still not : C:\Users\<name>\Manuscrit_pages>kraken -i page_1.png page_1.txt segment -bl ocr -m C:/Users/<name>/.local/share/kraken/models/FoNDUE-GD_v2_fr.mlmodel Loading ANN C:\Users\<name>\AppData\Local\Programs\Python\Python310\lib\site-packages\kraken\blla.mlmodel ✓ Loading ANN \Users\<name>\.local\share\kraken\models\FoNDUE-GD_v2_fr.mlmodel ✓ Segmenting page_1.png [12/05/25 13:34:48] WARNING Polygonizer failed on line 0: TopologyException: side location conflict segmentation.py:777 at 1101.6261767385363 272.87853021560886. This can occur if the input geometry is invalid. ✓ [12/05/25 13:34:50] ERROR Failed processing page_1.png: 'default' kraken.py:448 then I thought, perhaps it's because there is not enough text on page 1 because it's just mostly the title, so I tried page 2, C:\Users\<name>\Manuscrit_pages>kraken -i page_2.png page_2.txt segment -bl ocr -m C:/Users/<name>/.local/share/kraken/models/FoNDUE-GD_v2_fr.mlmodel Loading ANN C:\Users\<name>\AppData\Local\Programs\Python\Python310\lib\site-packages\kraken\blla.mlmodel ✓ Loading ANN \Users\<name>\.local\share\kraken\models\FoNDUE-GD_v2_fr.mlmodel ✓ Segmenting page_2.png ✓ [12/05/25 13:36:07] ERROR Failed processing page_2.png: 'default' kraken.py:448 it doesn't give exactly the same error, but comes from the same line...? -- Reply to this email directly or view it on GitHub: #749 (comment) You are receiving this because you are subscribed to this thread. Message ID: ***@***.***>

seems to go further, but not yet,

C:\Users<name>\Manuscrit_pages>kraken -i page_2.png page_2.txt segment -bl ocr -m C:\Users\\.local\share\kraken\models\FoNDUE-GD_v2_fr.mlmodel
Loading ANN C:\Users<name>\AppData\Local\Programs\Python\Python310\lib\site-packages\kraken\blla.mlmodel ✓
Loading ANN \Users<name>.local\share\kraken\models\FoNDUE-GD_v2_fr.mlmodel ✓
Segmenting page_2.png ✓
[12/05/25 14:08:22] ERROR Failed processing page_2.png: 'default' kraken.py:448

and this neither,

C:\Users<name>\Manuscrit_pages>kraken segment -i page_2.png -m C:\Users\\.local\share\kraken\models\FoNDUE-GD_v2_fr.mlmodel
[12/05/25 14:10:01] WARNING Baseline model (('page_2.png',)) given but legacy segmenter selected. Forcing kraken.py:522
to -bl.
Loading ANN page_2.png ✗

?

[PonteIneptique]

Please run with --raise-on-error after kraken

[mittagessen]

As I said the error is raised because the Windows path is parsed as a line-type model mapping. You need to find some way to escape the `C:\...` part, otherwise you're not going to get anywhere. Or just use Linux. That one is actually supported.

…

On 25/12/05 05:11AM, ytreza-azerty wrote: ytreza-azerty left a comment (mittagessen/kraken#749) seems to go further, but not yet, C:\Users\<name>\Manuscrit_pages>kraken -i page_2.png page_2.txt segment -bl ocr -m C:\\Users\\<name>\\.local\\share\\kraken\\models\\FoNDUE-GD_v2_fr.mlmodel Loading ANN C:\Users\<name>\AppData\Local\Programs\Python\Python310\lib\site-packages\kraken\blla.mlmodel ✓ Segmenting page_2.png ✓ [12/05/25 14:08:22] ERROR Failed processing page_2.png: 'default' kraken.py:448 and this neither, C:\Users\<name>\Manuscrit_pages>kraken segment -i page_2.png -m C:\\Users\\<name>\\.local\\share\\kraken\\models\\FoNDUE-GD_v2_fr.mlmodel [12/05/25 14:10:01] WARNING Baseline model (('page_2.png',)) given but legacy segmenter selected. Forcing kraken.py:522 to -bl. Loading ANN page_2.png ✗ ? -- Reply to this email directly or view it on GitHub: #749 (comment) You are receiving this because you commented. Message ID: ***@***.***>

C:\Users<name>\Manuscrit_pages>kraken --raise-on-error -i page_2.png page_2.txt segment -bl ocr -m C:\Users\\.local\share\kraken\models\FoNDUE-GD_v2_fr.mlmodel
Loading ANN C:\Users<name>\AppData\Local\Programs\Python\Python310\lib\site-packages\kraken\blla.mlmodel ✓
Loading ANN \Users<name>.local\share\kraken\models\FoNDUE-GD_v2_fr.mlmodel ✓
Segmenting page_2.png ✓
[12/05/25 14:29:10] ERROR Failed processing page_2.png: 'default' kraken.py:448
╭─────────────────────────────── Traceback (most recent call last) ────────────────────────────────╮
│ C:\Users<name>\AppData\Local\Programs\Python\Python310\lib\runpy.py:196 in │
│ _run_module_as_main │
│ │
│ 193 │ main_globals = sys.modules["main"].dict │
│ 194 │ if alter_argv: │
│ 195 │ │ sys.argv[0] = mod_spec.origin │
│ ❱ 196 │ return _run_code(code, main_globals, None, │
│ 197 │ │ │ │ │ "main", mod_spec) │
│ 198 │
│ 199 def run_module(mod_name, init_globals=None, │
│ │
│ C:\Users<name>\AppData\Local\Programs\Python\Python310\lib\runpy.py:86 in _run_code │
│ │
│ 83 │ │ │ │ │ loader = loader, │
│ 84 │ │ │ │ │ package = pkg_name, │
│ 85 │ │ │ │ │ spec = mod_spec) │
│ ❱ 86 │ exec(code, run_globals) │
│ 87 │ return run_globals │
│ 88 │
│ 89 def _run_module_code(code, init_globals=None, │
│ │
│ in :6 │
│ │
│ 3 if name == 'main': │
│ 4 │ if sys.argv[0].endswith('.exe'): │
│ 5 │ │ sys.argv[0] = sys.argv[0][:-4] │
│ ❱ 6 │ sys.exit(cli()) │
│ 7 │
│ │
│ C:\Users<name>\AppData\Local\Programs\Python\Python310\lib\site-packages\click\core.py:1442 │
│ in call │
│ │
│ C:\Users<name>\AppData\Local\Programs\Python\Python310\lib\site-packages\click\core.py:1363 │
│ in main │
│ │
│ C:\Users<name>\AppData\Local\Programs\Python\Python310\lib\site-packages\click\core.py:1862 │
│ in invoke │
│ │
│ C:\Users<name>\AppData\Local\Programs\Python\Python310\lib\site-packages\click\core.py:1799 │
│ in _process_result │
│ │
│ C:\Users<name>\AppData\Local\Programs\Python\Python310\lib\site-packages\click\core.py:794 in │
│ invoke │
│ │
│ C:\Users<name>\AppData\Local\Programs\Python\Python310\lib\site-packages\kraken\kraken.py:446 │
│ in process_pipeline │
│ │
│ 443 │ │ │ │ if len(fc) - 2 == idx: │
│ 444 │ │ │ │ │ ctx.meta['last_process'] = True │
│ 445 │ │ │ │ with threadpool_limits(limits=ctx.meta['threads']): │
│ ❱ 446 │ │ │ │ │ task(input=input, output=output) │
│ 447 │ │ except Exception as e: │
│ 448 │ │ │ logger.error(f'Failed processing {io_pair[0]}: {str(e)}') │
│ 449 │ │ │ if ctx.meta['raise_failed']: │
│ │
│ C:\Users<name>\AppData\Local\Programs\Python\Python310\lib\site-packages\kraken\kraken.py:245 │
│ in recognizer │
│ │
│ 242 │ │ │ │ │ │ │ tags_ignore=tags_ignore, │
│ 243 │ │ │ │ │ │ │ no_legacy_polygons=ctx.meta['no_legacy_polygons']) │
│ 244 │ else: │
│ ❱ 245 │ │ it = rpred.rpred(model['default'], im, bounds, pad, │
│ 246 │ │ │ │ │ │ bidi_reordering=bidi_reordering, │
│ 247 │ │ │ │ │ │ no_legacy_polygons=ctx.meta['no_legacy_polygons']) │
│ 248 │
╰──────────────────────────────────────────────────────────────────────────────────────────────────╯
KeyError: 'default'

[johnlockejrr]

C:\Users\Manuscrit_pages>kraken --raise-on-error -i page_2.png page_2.txt segment -bl ocr -m C:\Users\.local\share\kraken\models\FoNDUE-GD_v2_fr.mlmodel Loading ANN C:\Users\AppData\Local\Programs\Python\Python310\lib\site-packages\kraken\blla.mlmodel ✓ Loading ANN \Users.local\share\kraken\models\FoNDUE-GD_v2_fr.mlmodel ✓ Segmenting page_2.png ✓ [12/05/25 14:29:10] ERROR Failed processing page_2.png: 'default' kraken.py:448 ╭─────────────────────────────── Traceback (most recent call last) ────────────────────────────────╮ │ C:\Users\AppData\Local\Programs\Python\Python310\lib\runpy.py:196 in │ │ _run_module_as_main │ │ │ │ 193 │ main_globals = sys.modules["main"].dict │ │ 194 │ if alter_argv: │ │ 195 │ │ sys.argv[0] = mod_spec.origin │ │ ❱ 196 │ return _run_code(code, main_globals, None, │ │ 197 │ │ │ │ │ "main", mod_spec) │ │ 198 │ │ 199 def run_module(mod_name, init_globals=None, │ │ │ │ C:\Users\AppData\Local\Programs\Python\Python310\lib\runpy.py:86 in _run_code │ │ │ │ 83 │ │ │ │ │ loader = loader, │ │ 84 │ │ │ │ │ package = pkg_name, │ │ 85 │ │ │ │ │ spec = mod_spec) │ │ ❱ 86 │ exec(code, run_globals) │ │ 87 │ return run_globals │ │ 88 │ │ 89 def _run_module_code(code, init_globals=None, │ │ │ │ in :6 │ │ │ │ 3 if name == 'main': │ │ 4 │ if sys.argv[0].endswith('.exe'): │ │ 5 │ │ sys.argv[0] = sys.argv[0][:-4] │ │ ❱ 6 │ sys.exit(cli()) │ │ 7 │ │ │ │ C:\Users\AppData\Local\Programs\Python\Python310\lib\site-packages\click\core.py:1442 │ │ in call │ │ │ │ C:\Users\AppData\Local\Programs\Python\Python310\lib\site-packages\click\core.py:1363 │ │ in main │ │ │ │ C:\Users\AppData\Local\Programs\Python\Python310\lib\site-packages\click\core.py:1862 │ │ in invoke │ │ │ │ C:\Users\AppData\Local\Programs\Python\Python310\lib\site-packages\click\core.py:1799 │ │ in _process_result │ │ │ │ C:\Users\AppData\Local\Programs\Python\Python310\lib\site-packages\click\core.py:794 in │ │ invoke │ │ │ │ C:\Users\AppData\Local\Programs\Python\Python310\lib\site-packages\kraken\kraken.py:446 │ │ in process_pipeline │ │ │ │ 443 │ │ │ │ if len(fc) - 2 == idx: │ │ 444 │ │ │ │ │ ctx.meta['last_process'] = True │ │ 445 │ │ │ │ with threadpool_limits(limits=ctx.meta['threads']): │ │ ❱ 446 │ │ │ │ │ task(input=input, output=output) │ │ 447 │ │ except Exception as e: │ │ 448 │ │ │ logger.error(f'Failed processing {io_pair[0]}: {str(e)}') │ │ 449 │ │ │ if ctx.meta['raise_failed']: │ │ │ │ C:\Users\AppData\Local\Programs\Python\Python310\lib\site-packages\kraken\kraken.py:245 │ │ in recognizer │ │ │ │ 242 │ │ │ │ │ │ │ tags_ignore=tags_ignore, │ │ 243 │ │ │ │ │ │ │ no_legacy_polygons=ctx.meta['no_legacy_polygons']) │ │ 244 │ else: │ │ ❱ 245 │ │ it = rpred.rpred(model['default'], im, bounds, pad, │ │ 246 │ │ │ │ │ │ bidi_reordering=bidi_reordering, │ │ 247 │ │ │ │ │ │ no_legacy_polygons=ctx.meta['no_legacy_polygons']) │ │ 248 │ ╰──────────────────────────────────────────────────────────────────────────────────────────────────╯ KeyError: 'default'

Can you actually read the replies? And maybe the docs?

Windows is not supported

if I had a linux I would do it my dear

[PonteIneptique]

Dear @ytreza-azerty , people here are trying to help you, passive aggressiveness is not gonna get you anywhere.

would anyone know how to continue with this
KeyError: 'default'
?

ok, I though again about your constructive answers above, and found a solution, not do absolute path... avoiding the : that way,

C:\Users<name>\Manuscrit_pages>kraken --raise-on-error -i page_2.png page_2.txt segment -bl ocr -m "..\.local\share\kraken\models\FoNDUE-GD_v2_fr.mlmodel"
Loading ANN C:\Users<name>\AppData\Local\Programs\Python\Python310\lib\site-packages\kraken\blla.mlmodel ✓
Loading ANN ..\.local\share\kraken\models\FoNDUE-GD_v2_fr.mlmodel ✓
Segmenting page_2.png ✓
Processing ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% 20/20 0:00:00 0:00:04
Writing recognition results for page_2.png ✓

now,
would anyone have a way to make a loop on all 154 pictures, with that line above, instead of doing it by hand?