Form Recognizer : Draw bounding box with fill color on hand written pdf (on text (signature) in pdf) using python

Question

Form Recognizer : Draw bounding box with fill color on hand written pdf (on text (signature) in pdf) using python

Baskar Kumar 1

Hi Team,

we are trying to draw bounding box with fill color on hand written pdf on signature through coordinates extracted from Form Recognizer if its readable or modifying the extracted coordinates of text from below the signature and drawing bounding box on signature, so tried one hand written pdf converting into image. But after converting to image, text below signature or some other text is not properly extracted after again passing the converted image to Form Recognizer. so, not able to draw bounding box on converted image.

our scenario is to draw bounding box with fill color on signature in hand written pdf file where ever it appears in the pdf (in 50 pages pdf file appears in 7 pages )

Please let us know your valuable suggestion on this.

YutongTie-MSFT 53,966 Reputation points Moderator

2022-05-27T02:44:22.547+00:00

Hello @Baskar Kumar

Thanks for reaching out to us. I can understand your scenario, to make things clarify, the issue currently for your project is - after transfering the PDF to image, some text which can be read in PDF can not be read in the imgae now. Is that correct?

Could you please share how you transfer the image? And what kind of image you are transfering to? Sometime low quality tranfering will cause information loss.

Looing forward to your response.

Regards,
Yutong
Baskar Kumar 1 Reputation point

2022-05-27T06:20:47.383+00:00

Hi Yutong,

Thank you for your response.

Below is the python code which i used to convert pdf to image in png :

import fitz
zoom_x = 2.0 # horizontal zoom
zoom_y = 2.0 # vertical zoom
mat = fitz.Matrix(zoom_x, zoom_y) # zoom factor 2 in each dimension

filename = "Signature.pdf"
doc = fitz.open(filename) # open document
for page in doc: # iterate through the pages
pix = page.get_pixmap(mat) # render page to an image
pix.save("page-%i.png" % page.number) # store image as a PNG

Also can you please let me know :
1> are there any azure API to draw bounding box (on extracted signature) on handwritten pdf ?
2> if point1 is not there, are there azure API to convert pdf to image and image to pdf without information loss ?, since if i have to convert pdf to image to draw bounding box, again i have to convert the image with bounding box to pdf without information loss

Regards,
Baskar Kumar
YutongTie-MSFT 53,966 Reputation points Moderator

2022-05-27T09:58:45.147+00:00

Hello @Baskar Kumar

I think Computer Vision can make it but I will try on my end first and let you know.

Regards,
Yutong
YutongTie-MSFT 53,966 Reputation points Moderator

2022-05-31T17:18:13.02+00:00

Hello @Baskar Kumar

Seems like Form Recognizer is best choice for bounding signature - https://learn.microsoft.com/en-us/azure/applied-ai-services/form-recognizer/concept-custom#try-signature-detection

For transfer the PDF, have tried bounding first then converting it to image to avoid the loss?

For avoiding the loss, I don't think there is a product in Azure supports but I have find some internet resource for you : https://www.quora.com/How-do-you-save-a-PDF-as-a-JPEG-without-losing-quality

I hope this helps.

Regards,
Yutong
Baskar Kumar 1 Reputation point

2022-06-01T10:08:04.247+00:00

Hi Yutong,

Thank you for your suggestion and links

i tried custom template model for recognizing signature. If I have 2 page pdf document with signature and test with this custom template model, it detects signature only in first page. If signature is present in second page, it doesn't detect.

for example : If I have 10 pages pdf document with signature and test with this custom template model, it detects signature only in first page. If signature is present in remaining pages, it doesn't detect.

Can you please suggest how to handle it

Regards,
Baskar Kumar
YutongTie-MSFT 53,966 Reputation points Moderator

2022-06-02T00:48:46.397+00:00

Hello @Baskar Kumar

I can understand that, but this seems a bug, how about to raise a support ticket and assign a support engineer online with you to see what's the problem?

Please let me know if you have no support plan and I am willing to enable you a free one for this issue.

Regards,
Yutong

Your answer

YutongTie-MSFT 53,966 Reputation points Moderator

2022-05-27T02:44:22.547+00:00

Hello @Baskar Kumar

Thanks for reaching out to us. I can understand your scenario, to make things clarify, the issue currently for your project is - after transfering the PDF to image, some text which can be read in PDF can not be read in the imgae now. Is that correct?

Could you please share how you transfer the image? And what kind of image you are transfering to? Sometime low quality tranfering will cause information loss.

Looing forward to your response.

Regards,
Yutong
Baskar Kumar 1 Reputation point

2022-05-27T06:20:47.383+00:00

Hi Yutong,

Thank you for your response.

Below is the python code which i used to convert pdf to image in png :

import fitz
zoom_x = 2.0 # horizontal zoom
zoom_y = 2.0 # vertical zoom
mat = fitz.Matrix(zoom_x, zoom_y) # zoom factor 2 in each dimension

filename = "Signature.pdf"
doc = fitz.open(filename) # open document
for page in doc: # iterate through the pages
pix = page.get_pixmap(mat) # render page to an image
pix.save("page-%i.png" % page.number) # store image as a PNG

Also can you please let me know :
1> are there any azure API to draw bounding box (on extracted signature) on handwritten pdf ?
2> if point1 is not there, are there azure API to convert pdf to image and image to pdf without information loss ?, since if i have to convert pdf to image to draw bounding box, again i have to convert the image with bounding box to pdf without information loss

Regards,
Baskar Kumar
YutongTie-MSFT 53,966 Reputation points Moderator

2022-05-27T09:58:45.147+00:00

Hello @Baskar Kumar

I think Computer Vision can make it but I will try on my end first and let you know.

Regards,
Yutong
YutongTie-MSFT 53,966 Reputation points Moderator

2022-05-31T17:18:13.02+00:00

Hello @Baskar Kumar

Seems like Form Recognizer is best choice for bounding signature - https://learn.microsoft.com/en-us/azure/applied-ai-services/form-recognizer/concept-custom#try-signature-detection

For transfer the PDF, have tried bounding first then converting it to image to avoid the loss?

For avoiding the loss, I don't think there is a product in Azure supports but I have find some internet resource for you : https://www.quora.com/How-do-you-save-a-PDF-as-a-JPEG-without-losing-quality

I hope this helps.

Regards,
Yutong
Baskar Kumar 1 Reputation point

2022-06-01T10:08:04.247+00:00

Hi Yutong,

Thank you for your suggestion and links

i tried custom template model for recognizing signature. If I have 2 page pdf document with signature and test with this custom template model, it detects signature only in first page. If signature is present in second page, it doesn't detect.

for example : If I have 10 pages pdf document with signature and test with this custom template model, it detects signature only in first page. If signature is present in remaining pages, it doesn't detect.

Can you please suggest how to handle it

Regards,
Baskar Kumar
YutongTie-MSFT 53,966 Reputation points Moderator

2022-06-02T00:48:46.397+00:00

Hello @Baskar Kumar

I can understand that, but this seems a bug, how about to raise a support ticket and assign a support engineer online with you to see what's the problem?

Please let me know if you have no support plan and I am willing to enable you a free one for this issue.

Regards,
Yutong

Share via

Form Recognizer : Draw bounding box with fill color on hand written pdf (on text (signature) in pdf) using python

Your answer