Thursday, October 29, 2020

Getting Started with Salesforce Einstein Vision OCR (tabular model) - Part 2

In the previous Einstein Vision OCR blog post, we took a look at setting up Einstein Vision and viewed a sample usage with Business card. Click here if you missed it. 

In this post, we shall see the other version of the OCR which is reading tabular format image. E.g Balance sheet, excel screen capture, inventory list, etc...

Official Einstein Vision Tabular documentation here:

I have created a sample code which will call the tabular model and from the predicted result, I built a html table which is being displayed right away in your Salesforce.

Demo screenshot

Here is a demo video:

Technical details about the prediction result: 

  • the recognized label will contain a cellLocation with rowIndex and colIndex. 
  • use the rowIndex and colIndex to map it on a two dimensional array 
  • then build the html table with tr and td
    • tr - for each row
    • td - for each column

Here are some use cases where it can be useful:
  • You need to digitalize all paper work tabular transaction, just scan it and let Einstein do the rest.
  • scan you sheets and display form for a data entry to validate and correct where necessary. This will save time during data entry

In the next post, we shall talk about the Einstein Vision Retail Execution API. Stay tuned.


No comments:

Post a Comment