Open chreul opened 7 years ago
Hi,
I just discovered OCR4all and first let me say thank you for all the effort and for making this available. It is probably exactly what I need to OCR some old typed indexes of what you call Kirchenbücher.
(I've also subscribed to the mailing list but it's not approved yet so I'm posting it here as it might be relevant for others landing on this same github issue..)
I am going through all the different steps in the user guide and have now arrived at the segmentation stage with Larex
As this issue is still open, I understand that different RoI for odd/even pages is not implemented yet. I was wondering if you have any tips/workarounds as I am facing this problem right now, where the header on the even pages is lower in the image than the header on the odd pages..
I could look into doing some extra preprocessing with ScanTailor, but as I have quite a bit of pages to process, it would of course be easier if you know of a workaround..
Best regards,
birt
Hi,
I found the odd/even selection option in Batch Segmentation, so I should be able to use that as a workaround.
So never mind my question :-)
Greets,
Birt
I'm happy to hear that the odd/even selector in the batch segmentation modal worked as fallback option but I guess specific options for e.g. RoI for recto/verso pages would probably still be a bit more elegant for a lot of use cases. We're definitely still planning to implement this in the future but as we're currently working on remodeling other core components with higher priority this may still take some time.
It should be possible to define two separate RoIs recto/verso (odd/even) pages.