Document Intelligence Incorrectly Rotating Table

honeymoow 0

I'm trying to OCR some tables in PDF form. When the column headers are vertical text, the output table is OCR'd correctly but rotated so that the column headers run up and down the right-most column (almost like a 90 degree rotation of the table). This happens sometimes and not other times. How can I stop this behavior? I'm using the Python API but it also happens on the online demo.

VasaviLankipalle-MSFT 15,836 Reputation points

2024-06-18T04:34:43.65+00:00

hello @honeymoow , Thanks for using Microsoft Q&A Platform.

Is it possible to share more details like the model and it's API version?
honeymoow 0 Reputation points

2024-06-18T12:08:11.5666667+00:00

@VasaviLankipalle-MSFT Document Intelligence Studio, Layout extraction, API version 2024-02-29.
VasaviLankipalle-MSFT 15,836 Reputation points

2024-06-19T21:53:51.14+00:00

@honeymoow , is it possible to share the sample document here or in the private message as requested to reproduce the same on our end?
honeymoow 0 Reputation points

2024-06-20T14:08:46.6333333+00:00

@VasaviLankipalle-MSFT Sure, I've attached several example documents that get incorrectly rotated so that the column headers are on the right and the table cells read from right to left (instead of top to bottom). I was unable to train a custom model that would redress the issue, either. Example A: haryana_2019_GANAUR.pdf; Example B: haryana_2019_ADAMPUR.pdf.
honeymoow 0 Reputation points

2024-06-24T13:47:45.4966667+00:00

@VasaviLankipalle-MSFT do you have an update on this?

Share via

Document Intelligence Incorrectly Rotating Table