Improve the result of OCR with alfresco-simple-ocr

[imanez1]

Hello,

I'm using the Alfresco addons alfresco-simple-ocr with pdfsandwich to extract data from an invoice, everything works fine but the results are not very accurate.

my invoice has this template:

and the result that i get after OCR is this:

BILLING ADDRESS INVOICE
XXXX XXXX
XXX XXXX XXXX Number 545614513
XXXX XXXX Date May 30, 2019
XXXX XXXX
Delivery No. INV1254
DELIVERY ADDRESS Your Request Date 	May 30, 2019
XXXX XXXX
XXX XXXX XXXX Your Order No. SO655614
XXXX XXXX Contract No. -
XXXX XXXX Quote No. SO655614
Customer No.
Your Contact 	XXXXX
(152)-568-5458
Our Contact XXXXX
admin@yourcompany.example.com

Pos. Prod.No. Description Qty Price/Item (USD) VAT Total (USD)
1 P_21154	XXXX 1 0.20 15% 0.20
Total USD (excl. taxes) 0.20
VAT 0.03
Total Net Price in USD (incl. VAT) 0.23

So my questions are:

- How can I improve the accuracy of the results? Because sometimes for example: instead of an 'S' I get a '5' or '8' instead of '8'.... 

- How can get the results in blocks, part by part: Part1, part 2 and then part 3

I tried croping the invoice with this command line and it gives me the results i want, but how can i do it from inside Alfresco?:

convert -density 200 INVOICE.pdf -crop 100x50% +repage \( -clone 0 -crop 50x100% +repage -reverse \) -delete 0 -reverse INVOICE-out.pdf

alfresco-global.properties:

# OCR #

ocr.command=/usr/bin/pdfsandwich
ocr.output.verbose=true
ocr.output.file.prefix.command=-o

ocr.extra.commands=-verbose -rgb -lang fra+eng -nopreproc 
ocr.server.os=linux

[angelborroy]

Try OCRmyPDF, it will give you better results.

https://ocrmypdf.readthedocs.io/en/latest/

[imanez1]

OCRmyPDF gives me more or less the same results.

How can i OCR block by block (part1, part2, part3)?

[anuradha1]

i am having the same issue. Did u found any solution?

[imanez1]

I'm still facing the same issue.

Anyone has an idea on how to extract the text from the invoice by parts?