Extract Text with OCR

(0) Share

Report

Posted on by Community Power Pla...

Hi all,

I would like to extract text from a website login page after entering the password.

This text is dynamic when the website is loaded each time.

Is there a way to identify the specific subregion of text that the OCR needs to read?

Regards

Hidayat

Categories:

Power Automate Desktop

I have the same question (0)

All responses (11)

Answers (1)

DanielOlsson 111 on at

Like (1)

Report

On action Extract text with OCR you can choose "Search mode" and add X and Y cords of the area.

Was this reply helpful? Yes No
Community Power Pla... on at

Like (0)

Report

Hi @DanielOlsson ,

Thanks for the feedback.
I tried this but received an error, Failed to extract text with OCR.

For the image to select, I tried both with the text and without text.

Image without text to capture

Image with text to capture

Regards
Hidayat

Was this reply helpful? Yes No
Verified answer

geavgous Microsoft Employee on at

Like (2)

Report

Hi Hidayat,

Thanks for exploring the potential of Power Automate Desktop! Instead of OCR, you could use the Web Automation action , called "Extract Data from Web". This action allows you to select specific web element and get anything you want from it.
You can learn more by checking the documentation page in Web automation - Power Automate | Microsoft Docs.

In addtion, you could use the Image based recording for using OCR. You can find more details in Recording in a desktop flow - Power Automate | Microsoft Docs.

Let me know how it goes!

Thanks,
George

Was this reply helpful? Yes No
MichaelAnnis 5,727 Moderator on at

Like (1)

Report

I like using the "Create Tesseract OCR engine" and using the multipliers before using "Extract text with OCR"
Mess with the multipliers until you get the result you need. The higher the multiplier, the slower the bot; however, it greatly increases the accuracy of the OCR.

Best of luck.

Was this reply helpful? Yes No
Community Power Pla... on at

Like (0)

Report

Hi @DanielOlsson, @geavgous , @MichaelAnnis ,
Thank you for pointing me into the intended direction.
I manage to get the solution by combining all of you guys feedback 😀
And reading the documentation should instead be the first step. 😅
Awesome work all!

Overview of Workflow
Extract text with OCR step
1) Search mode: Selected subregion relative to image
2) Tolerance: Increased to 10

Regards
Hidayat

Was this reply helpful? Yes No
JAWL 393 on at

Like (0)

Report

Hi Hidayat,

1. How you actually identify the X1, X2, Y1, Y2 coordinates? Can we capture the coordinates from the position of mouse?
2. I use Capture Image in the OCR action but it's display is blurred. Or you upload clear image instead?

Was this reply helpful? Yes No
DanielOlsson 111 on at

Like (0)

Report

Hello, you can use the move mouse action to determine the X and Y cords of your picture. Not as an active action but while you develop your flow. Your picture might be blurred as its zooms in when it take the snapshot but the quality is as well determined by the resolution of your screen as well as how you connected to the device you run your flow on, is it local or do you make some kind of remote connection?

Was this reply helpful? Yes No
Salman222711 4 on at

Like (1)

Report

Hello Jawl, you can use move mouse to image activity>Advanced>Search Mode>Search on Specified subregion of screen or foreground Window.Then you will be getting option to select the region.By double tapping and dragging the area X1,X2,Y1,Y2 Positions will be reflecting.These positions you can note down and use in other OCR actions.Please Check the below image.

Thanks Regards
Salman

Was this reply helpful? Yes No
JAWL 393 on at

Like (0)

Report

Is it possible to use OCR action move mouse to text found based on the whole multipage pdf file (and not the active screen)?

Was this reply helpful? Yes No
RRRprogram 5 on at

Like (0)

Report

Excellennttt brother...Since I am non Computer science background.. I was struggling to find this answer... thanks for your help.

Was this reply helpful? Yes No