Handwritten signature Recognition

Help
1

Hi,

I want to identify signature is present is multiple scanned image or pdf files in a folder. I tried to use image exists activity inside for loop. But it is not recognizing the signature. Is there any other methods to fullfill this task? Can anyone help me on this please?

Thanks in advance!

Regards,
Renju

1 Like
2

You can’t do that al least for the moment :slight_smile: . This is called machine learning because it is almost imposible to sign every time perfectly identical and you need many samples with that signature. I am sure that UiPath and other OCR companies are working at this chapter “handwrite” and we are waiting to amaze us again :slight_smile:

3

Hi,

Thanks for the update @VladTof. Is there any possibilty to identify the presence of such characters. I don’t to extract the data, I just need to identify whether signature is present in each file.

Regards,
Renju

4

It is the same thing because the tricky part is to identify it.
I my opinion the only solution is with image exists but it is not applicable because the signature will not be the same twice.

5

Hi @renjutom,

image
I’m not an expert in OCR but I think you can use ‘Get OCR Text’ for this to work correctly, I recommend using anchor to a text that is always present for example:
image
In the above image your anchor will be the word ‘signature’ since it will always be present and you will try to extract what’s in the box. As long as what’s in the box is not empty then that means there’s a signature right? I don’t know how reliable this might be but I suggest you to try it. Let me know if you were able to archive this. There are multiple OCR tutorials in academy.uipath.com it might help as well, good luck!

2 Likes
6

You can do what you are asking. But, you will still need some secondary validation from a human before sending it off to the next step.

First, you need to get the box or element around the signature. To do so, you need to find the box… So if the pdf is an image where you can’t highlight any text, then you must use the Find Image of text near the signature box, while scrolling each page until you find the image (preferably starting from the bottom, since the signature is usually near the end). If the pdf is not an image, then you can use Find Text or Find Element; in this situation, you may even be able to find the signature box element. You can also use Anchor Base, however I can’t confirm it’s effectiveness when the element is a signature box.

After, you have the element of the signature box or the element near the signature box, you will want to use the Set Clipping Region to adjust the surrounding box to only include the signature area (use the Highlight activity to help visually see your adjusted box until it is correct). Do some searches on setting the clipping region for syntax on this, because it has been a while since I’ve done this.

Now, that you have a clipping region stored in an element variable, you can then use Get OCR Text using that element. You can even use Save Image, if you would like to insert the image into Excel for the secondary validation. To validate the signature with the bot, you just need to set an array of characters/strings which are identified as invalid such as “”,“.”,“,”,“/”, etc, then use a condition like IF invalidCharacters.Any(Function(x) x = sigText.Trim), which will be True or False depending on if the OCR text of the signature box has enough there to identify that “something” is there. Note: you will need to play around with the Scale for the OCR engine so it is more accurate.

Anyway, just sharing my experience with this, and hope it helps.

Regards.

7 Likes
7

Here is one of my old posts with dealing with scanned pdfs:
https://forum.uipath.com/t/checkbox-and-signature-on-scanned-pdf-any-ocr-engine-in-uipath/8140/2?u=claytonm

8

Hi @ClaytonM and @rmunro ,

Thanks for the suggestions!

I will try with the above methods and update the status.

Regards,
Renju

1 Like
9

thanks a lot

10

@renjutom,
I am trying to achieve the same thing. Touching base to see if you had any luck ?