How to train UIpath to solve captcha image?

Help Activities
, ,
1

Hi everyone,

Please give some help, i need to get the text of captcha images to perform an arithmetic operation. I need to use the components of community edition of UiPath. Can you suggest me the best way to solve it? I’ve tried Document Understanding following the Academy steps but no results.

Any help would be appreciated.
Thanks.
Larry

  • Here some source examples:
    2-1
2

You can’t. That’s the POINT to Captcha images, to stop robots.

There are various web sites you can find that are dedicated to solving them for you and returning the result.

3

Hey,

You could try the below thread:

However captchas keep changing and it will always be a pain.

In our solution we did this manually and then bot use to pick it up using triggers after that.

4

Hi @larrysoft

Extracting text from CAPTCHA images can be challenging due to the deliberate obfuscation techniques used in CAPTCHAs to prevent automated recognition.

There are third-party services available that specialize in solving CAPTCHA challenges. These services often provide APIs that you can integrate into your UiPath automation. You can send the CAPTCHA image to the service, and they will return the recognized text. However, note that these services may come at a cost.

Thanks!!

5

Hi @larrysoft

  1. Use Get OCR Text to capture the captcha and store it in an variable.
  2. Use Find Matching Patterns to separate both the integer and store it in an variable. Use the following regex expression to separate both integers:
First Integer: [\d]+(?=\+)
Second Integer: (?<=\+)[\d]+
  1. Use an Assign activity and convert the string to integer and follow the syntax:
Output: CInt(Match1)+CInt(Match2)
  1. Print the output variable in log message or writeline.

Note: Attached the workflow for the reference.
Sequence5.xaml (16.2 KB)

Hope it helps!!
Regards,

6

Hi,

Included OCR activities didn’t resolve captcha due to obfuscation applied to the image as you can see in the example.

I’m trying to understand cognitive services from azure in order to validate the image but it implies training a model.

Any help would be appreciated.

Thanks.