hellostream 7 - afsug.com · about eric kreft eric kreft ca(sa), senior manager applications at...
TRANSCRIPT
hellostream 7
hello Eric Kreft & Pieter ConradiePython Imaging and OCR with SAP ERP - Fact or Fiction
AfriSam South Africa
About Eric KreftEric Kreft CA(SA), Senior Manager Applications at AfriSam, with over 20 years experience in IT having implemented SAP and related products literally from A(Afaria) to Z(Z transactions) across multiple business units, countries and modules.
My passion is to keep things simple and sustainable and use SAP in it’s natural “environment” to the greatest extent possible.
the problem
Approximately 3000 deliveries per day
POD (Proof of Delivery) signed by customer required
Lots of paper documents to be scanned and matched
to SAP delivery note in SAP
existing solution
Third party software to process scanned documents
Scanned document linked to SAP using RFC and SAP archivelink
functionality from the software
Users use software to handle errors related to documents not being
linked automatically in software
Volumes that required manual intervention quite high
Cost of licence linked to image pages scanned
the options
Keep existing software solution, major cost implication due
to license model and processing backlog keeps growing
Find another solution, but needs to be Simple, Sustainable
and “Part of SAP” !!
the solution
Python !!! – But Why ? Lots of packages to do just about anything (https://pypi.org)
Runs on existing infrastructure (SAP TREX is python based)
No impact on existing hardware (Used the TREX server)
Minimal additional packages needed (Some additional packages loaded to standard SUSE SAP Install)
Clean integration into SAP using SAP NW RFC SDK (PyRFC)
Has the latest OCR options – Tesseract (Google) with neural networks, open source and supported by
the features
SAP ERP owns the process !!
SAP “BOT” to run the process
“automatically”
manual work
Image quality as scanned at plant locations
OCR is not the human eye !!
process efficiency
How to handle approx. 30 – 40 % of manual images – 900 per day
Simple transaction for user, “BOT” does the rest
Full size
Cropped “OCR” image
Manual entry transaction
lessons
OCR is not an exact science !!!
Image quality always an issue
Provide for manual intervention – but must be efficient
Build your own “BOT” in SAP
There must be an owner
Pilot with iterations
Be open to latest tech – SAP with Python?
Keep it SIMPLE.. – sometimes use ‘IBM’ !
interesting info
https://stackabuse.com/pytesseract-simple-python-optical-character-recognition/
https://github.com/SAP/PyRFC
https://pypi.org/
About Pieter ConradiePieter is a Business Applications Consultant with more than 22 years’ experience in the SAP environment. Pieter is SAP certified in both Materials Management Module as well as the Business Warehouse Application. He cut his teeth in the Plant Maintenance environment and is a self-taught ABAP programmer, developing numerous custom applications and solutions whilst employed with AfriSam. He is an avid DIYer and this passion spills over into providing solutions to the Afrisamcommunity. Pieter is currently working on a delivery execution monitoring system in the outbound logistics space, with a focus on improving on time delivery to the AfriSam customers.
the end goal
Link vendor invoice images to
SAP Invoice Document
The experiences with 3rd party
provided scanning solutions
Cost effective solution for
subsidiary companies
the road to success
Provide functionality to load files into
application server file system
OCR tool to split files into individual pages
Perform OCR
Save result to database for processing
the road block
Complex documents can
generate numerous lines of text
At face value, making sense of
extracted data perceived to be
challenging
the shortcut
Regex (Regular Expressions)
SAP’s DEMO_REGEX_TOY
Extract meaningful data to
identify supplier
the direction markers along the road
OCR at times not accurate
when extracting data within
complex invoices
Extracting document parts
proved accuracy could be
enhanced
Introduce cropping functionality
Regex ruleset developed
the direction markers along the road
OCR at times not accurate
when extracting data within
complex invoices
Extracting document parts
proved accuracy could be
enhanced
Introduce cropping functionality
Regex ruleset developed
the direction markers along the road
OCR at times not accurate when
extracting data within complex
invoices
Extracting document parts
proved accuracy could be
enhanced
Introduce cropping functionality
Regex ruleset developed
the direction markers along the road
OCR at times not accurate when
extracting data within complex
invoices
Extracting document parts
proved accuracy could be
enhanced
Introduce cropping functionality
Regex ruleset developed
the well marked route
the arrival at destination
the arrival at destination
the arrival at destination
the potholes
Master data
Absent data on invoice images
Poor handling of paper documents
Poorly scanned images
Hand written images
reflecting on the gems along the road
The ability of SAP to integrate with external apps
The power of Regex:
https://en.wikipedia.org/wiki/Regular_expression
https://www.regular-expressions.info/
SAP GUI toolbox: ALV, Custom controls
the journey ahead
Extracting and attaching statement data to
the statement reconciliation application
Cropping and extracting image data whilst
validating by interacting with the stored
image
Rollout to all AfriSam Companies ;-)
contactdetails
Name : Eric Kreft / Pieter Conradie
Company : AfriSam South Africa
Email : [email protected] / [email protected]
Contact # : 011 670 5500
Website : https://www.afrisam.co.za
thankyou