Hello,
I have been processing the same type of documents for 3 years now with no issues. However, these last 3 days, these documents are being processed incorrectly and the text output is returning with a lot of errors (incorrectly recognized characters, numbers recognized as letters, etc.). It seems that something changed since these worked flawlessly always.
If there were any changes to the API or something, can you clear this out for us?
Regards,
Manuel (Citamed)
コメント
4件のコメント
Hello,
The last update of OCRT in Cloud OCR SDK was on February 28, 2018, so the issue cannot be connected with any changes on our side. Could you please send to CloudOCRSDK@abbyy.com some image samples for which the issue can be reproduced? Possibly you need to tune the processing settings for these images.
Hello, attached are two examples. The original one in PDF format and the OCR'ed one in TXT format.
https://www.dropbox.com/s/zvgus6dhuzgndiv/2018-7-16_4.pdf?dl=0
https://www.dropbox.com/s/7v1a984w4eion63/1531746627024.txt?dl=0
Hello!
Here are two more examples that happened today!
Source File: https://www.dropbox.com/s/74pe30ln13koeoi/2018-7-26_151.pdf?dl=0
OCR'ed File: https://www.dropbox.com/s/7v1a984w4eion63/1531746627024.txt?dl=0
Thanks!
Hello,
Thank you for the provided documents.
Please try to use the imageSource=scanner parameter of the processImage method for your documents. In this case some types of geometrical distortions is not corrected. As we understand, you mostly process PDF documents, and this specific image preprocessing option is intended mostly for photos and is not necessary to be used for scanned document of good quality. The auto mode sometimes can mistakenly correct the distortions when it is not necessary.
With this settings your documents have been processed more accurately.
サインインしてコメントを残してください。