Closed jbarth-ubhd closed 3 years ago
Thanks for the report!
Looks like this is related to this SO issue around this GH issue of Pillow. Which version do you have? (pip show pillow
)
BTW your workflow has a flaw: You cannot use clipping after any processor that already adds derived images (AlternativeImage
) on the same hierarchy level. Since...
...and so the processor won't actually do anything but print warnings...
I have updated the workflow guide wiki page to better reflect this. (I distinctly remember documenting this elsewhere already, though...)
EDIT So I would recommend exchanging the order with the region-level deskewing. Also, you don't need clipping on the line level if you already apply resegmentation. The latter can be seen as an alternative method (based on coordinates instead of image data). For full clarification about both operations including screenshots, I recommend these slides (and follow-up).
Thanks for the report!
Looks like this is related to this SO issue around this GH issue of Pillow. Which version do you have? (
pip show pillow
)
(venv) jb@pers109:/usr/local/ocrd_all> pip show pillow
Name: Pillow
Version: 7.2.0
Workflow has been generated by script according to https://ocr-d.de/en/workflows (as of ~ 2020-06).
You cannot use clipping after any processor that already adds derived images
this two lines?
cis-ocropy-deskew -I OCR-D-N7 -O OCR-D-N8 -P level-of-operation region" \
cis-ocropy-clip ...
But it's still here: https://ocr-d.de/en/workflows Step 9 + 10
You cannot use clipping after any processor that already adds derived images
this two lines?
cis-ocropy-deskew -I OCR-D-N7 -O OCR-D-N8 -P level-of-operation region" \ cis-ocropy-clip ...
But it's still here: https://ocr-d.de/en/workflows Step 9 + 10
Yes, sorry, did not spot it there the first time. I have updated my issue on the recommendations.
Thanks for the report! Looks like this is related to this SO issue around this GH issue of Pillow. Which version do you have? (
pip show pillow
)(venv) jb@pers109:/usr/local/ocrd_all> pip show pillow Name: Pillow Version: 7.2.0
7.2.0 should have the fix IIUC. I am still trying to reproduce. It definitely depends on the image's color mode (16/32 bit integer or float are probably the culprit). Getting back to you when I tried with such an image.
I am still trying to reproduce. It definitely depends on the image's color mode (16/32 bit integer or float are probably the culprit). Getting back to you when I tried with such an image.
No luck here. Tried with 8+1, with 16 bit and 32 bit images. Could you please share the image it failed on?
Sent an email with link to images.
Thanks – I could reproduce. You helped find a bug again! Fix is in #75
Workflow:
Log: