Improve the performance of vertical text recognition for Form Recognizer Read API

Question

Hello,

We have been working on a computer vision use case that involves performing OCR on images using Form Recognizer. The text that appears in the images includes both horizontal and vertical alphanumeric texts. We have tested the Form Recognizer Read API extensively and found it underperforms in vertical text detection and recognition, especially when the text is alphanumeric. It looks like the read models were not trained on alphanumeric vertical text; we have not yet seen it recognize one alphabet in all our vertical text samples.
Attaching a few samples for you reference.

Do you have any ideas on how we can improve the recognition of vertical text?

Answer 1

Hello @Kenrick Fernandes , welcome to Microsoft Q&A Platform.

Thank you for providing the detailed information and bringing this to our notice. I just reproduced using latest preview API version yes it looks like that the Form Recognizer Read API is not extracting accurately. Will share this feedback to the product team.

If your scenario is something related to extract labels, street signs and posters please use Azure AI Vision v4.0 preview Read feature as mentioned here. https://learn.microsoft.com/en-us/azure/ai-services/document-intelligence/concept-read?view=doc-intel-3.0.0

For extracting text from external images like labels, street signs, and posters, use the Azure AI Vision v4.0 preview Read feature optimized for general, non-document images with a performance-enhanced synchronous API that makes it easier to embed OCR in your user experience scenarios.Try custom model in FR studio and see if it helps.

I hope this helps.

Regards,
Vasavi

-Please kindly accept the answer and vote 'yes' if you feel helpful to support the community, thanks.