OCR Capabilities for scanned documents

[sdeokule]

Hello team,

In one use case I'm working on, data needs to be extracted from scanned PDF copies of all documents.

I have tried by using Extract text with OCR activity with two options like windows OCR engine and Tesseract OCR engine but it seems the accuracy is only 60% with these OCR engines.

Is there any other OCR engines which is suitable for scanned documents with more than 90% accuracy?

I am good if it is paid also.

[momlo]

Hi @sdeokule 

You can use AI Builder wich is available as part of Power Platform and as a cloud service. It costs money though.

https://powerautomate.microsoft.com/en-us/ai-builder/

[Pavel_NaNoi]

If you're paying for flows, you can get a free trial of AI builder and get 5k credits spare each month (which is about 10 documents iirc), use that to see if its what you need.

[sdeokule]

Hello Momlo,

Yeah there are options through AI builder but i am expecting these actions in power automate desktop because i have some other dependency.Is there any option through PAD to invoke these services

[momlo]

yes, you can do this in a several ways - you can search the forum and you will find posts where i described some of my flows

In a nutshell, three options:

- set up power automate flow triggered by api, then in PAD call this flow with web request, passign the file

- set up power automate flow the triggers for each new file in a sharepoitn library, then use sharepoint actions to save file to sharepoint

- set up power automate flow the triggers for each new file in a sharepoitn library, then save the file into onedrive fodler on the machie, that is synchronised with onedribe/sharepoint

In each scenario your power autoatme flows needs to:

- pass the file vai ai builder

- pass the results back to pad, if run via web call - as a response, if file stored on sharepoint - store in a excel or json or whatever file you wil lthen read back in PAD