By using AWS re:Post, you agree to the AWS re:Post Terms of Use
AWS re:Postre:Post

Textract wrong detection based on a opencv made table

0

A table and embedded original images Hello everyone, I made a table using opencv and put some images of 97 in each of the table's grids. The 97 image itself can be detected, and if I only put one 97 in the top left, it also can be detected. However, with the above table, all 97 will be detected as L6. I tried different size and distance but none seems to work. Can someone please provide me a hint or help? Many thanks in advance!

Topics
Machine Learning & AI
Tags
Amazon Textract
Language
English
ocrtest
asked 8 months ago195 views
1 Answer
0

Doesn't look like a real use case. The L6 seems to indicate that Textract identified the image to be rotated 180 degrees. You can verify that using: https://github.com/aws-samples/amazon-textract-response-parser/blob/master/src-python/README.md#page-orientation-in-degrees

You should get better accuracy with documents that resemble a more realistic use case.

AWS
Martin Schade _AWS_
answered 8 months ago
  • ocrtest
    8 months ago

    Hi Martin,

    Thank you for your quick response! Could you explain a bit more what it means by saying "Doesn't look like a real use case"? And what could be the reason and solution for the issue that Textract detects this table as 180 degrees rotated? Some other numbers I put in work well though. Thanks again!

You are not logged in. Log in to post an answer.

A good answer clearly answers the question and provides constructive feedback and encourages professional growth in the question asker.

Guidelines for Answering Questions

Relevant content