"OCR Extract" action doesn't work well (alfresco-simple-ocr + pdfsandwich)

[hisayo-s]

Hello,

I'm using Alfresco 5.2 community edition on CentOS7.5 and it works well itself.

Now I trying to add OCR function to Alfresco, so I installed alfresco-simple-ocr (simple-ocr-repo-2.3.1.jar) and pdfsandwich to add function.

When I install pdfsandwich version 1.4, ruled "Extract OCR" action do works and version 1.1 PDF-file made automatically. But all pages of OCR PDF are white paper; no images, no characters.

Secondly I uninstall pdfsandwich version 1.6 insted of version 1.4, and tried again. Then ruled "Extract OCR" action DO NOT seem to be occured, and version 1.1 PDF file never made.

I tried pdfsandwich version 1.4, 1.5, 1.6 and 1.7 on comannd-line, and they works well expect version 1.7. (Version 1.7 says buggy message on command line)  When use version 1.4, 1.5, 1.6, exit-code is zero. 

---------------------------------------------------------------------------

RULE DEFINITION

Attached file is screen shot of rule definition. (Japanese)

When item created or input on this folder OR when item updated,

AND MIME-type is "Adobe PDF Document",

execute "Extract OCR".

- Continue on error: Checked

- Execute the rule background: Checked

---------------------------------------------------------------------------

/opt/alfresco-community/tomcat/shared/classes/alfresco-global.properties

　:

　:
### Alfresco Simple OCR ###
ocr.command=/usr/local/bin/pdfsandwich
ocr.output.verbose=true
ocr.output.file.prefix.command=-o

ocr.extra.commands=-verbose -rgb -lang jpn
ocr.server.os=linux

---------------------------------------------------------------------------

Anyone please help me !

[fedorow · Accepted solution]

So, to make it works on Alfresco/Share CE 6.1.2-ga/6.1.0 I made shared volume between alfresco and ocrmypdf containers. I replace /ocr_input and /ocr_output to one directory /ocr and map it as volume for both containers. 

Only one problem, asynchronous mode for rule gives me error. So I turn it off.

Angel thanks!

docker-compose.yml

...
services:
   alfresco:
      ...
      volumes:
         - ocr:/ocr
      ...

   ocrmypdf:
      ...
      volumes:
         - ocr:/ocr
   ...
volumes:
   ...
   ocr:
      driver: local
...

bin/ocrmypdf.sh

(and remove {} from $OUTPUT_FILE_PARAM in copy output file command) 

#!/bin/bash

INPUT_DIR=/ocr
OUTPUT_DIR=/ocr

# ocrmypdf hostname
OCRMYPDF_SERVER="ocrmypdf"

# identify parameters, input and output file
array=( "$@" )
len=${#array[@]}
ARGS=${array[@]:0:$len-2}

LAST_ARGS="${@: -2}"
INPUT_FILE_PARAM=`echo "$LAST_ARGS" | cut -d ' ' -f 1`
OUTPUT_FILE_PARAM=`echo "$LAST_ARGS" | cut -d ' ' -f 2`

# extract filenames
INPUT_FILE=$(basename "$INPUT_FILE_PARAM")
OUTPUT_FILE=$(basename "$OUTPUT_FILE_PARAM")

# SSH parameters
SCP=cp
SSH=ssh
USER=root

# copy original pdf to ocrmypdf server
$SCP $INPUT_FILE_PARAM $INPUT_DIR

# execute ocrmypdf program
$SSH $USER@$OCRMYPDF_SERVER "/usr/bin/ocr.sh $ARGS $INPUT_DIR/$INPUT_FILE $OUTPUT_DIR/$OUTPUT_FILE"

# copy transformed pdf back to alfresco path
$SCP $OUTPUT_DIR/$OUTPUT_FILE $OUTPUT_FILE_PARAM

# remove temporal files
rm -f $INPUT_DIR/$INPUT_FILE $OUTPUT_DIR/$OUTPUT_FILE

[douglascrp]

Hello.

Check the following link FAQ · keensoft/alfresco-simple-ocr Wiki · GitHub 

Maybe that can help you.

[hisayo-s]

Hi,

Thank you for your information.

I haven't read the FAQ page, so I will read the FAQ carefully and try to improve my environment.

I hope good result.

Best regards,

[hisayo-s]

Hello,

My problems has been partly solved.

Reading FAQ, I installed 2 jar files (simple-ocr-repo-2.3.1.jar and simple-ocr-share-2.3.1.jar) insted of  simple-ocr-repo.amp. (I had used amp file)

After restarted alfresco, the "Extract PDF action" sometimes works well, and sometimes not.

When action(conversion) succeed, "tesseract" ".convers.b+" "unpaper" processes are running on "top" view.

Otherwise when action(conversion) fails, their processes appears shortly and soon disappears.

It seems that file size and number of page are unrelated. 

I have no idea how to solve this problem.

Anyone know the solution. Please let me know!

------

- CentOS 7.5

- Alfresco 5.2 - community edition

- alfresco-simple-ocr 2.3.1

- pdfsandwich is 1.6 (*1)

- tesseract 3.04

(*1)

When I tested version 1.7 again, pdfsandwich says following message as before.

> "Fatal error: exception Unix.Unix_error(Unix.ENOTEMPTY, "rmdir", "/tmp/pdfsandwich_tmp2d3ca3")"

Such being the case, I use version 1.7 with "-debug" option to avoid error. (temp files should be erased manually...)

[angelborroy]

Try using OCRmyPDF (https://github.com/jbarlow83/OCRmyPDF) instead of pdfsandwich.

Both pdfsandwich and OCRmyPDF have some issues on CentOS (they are developed for Ubuntu), but you can use the Docker Image for OCRmyPDF available at https://ocrmypdf.readthedocs.io/en/latest/installation.html#installing-the-docker-image