[Feature] Optical Character Reader [OCR] about omrchecker HOT 10 OPEN

Yes, we have a related issue to add a plugin for the same. Can you help me formulate the problem with more clarity?

We need help in generalizing the approach.

from omrchecker.

Hello, I can clearly contribute to this issue. As I have many a times worked upon tesseract OCR, Vision API OCR, and Microsoft OCR wherein I had made changes in the OCR by computer vision techniques like gray scaling the image and even giving it proper contrast to read the text accurately.

from omrchecker.

Thank you Adwait01. i am also working on this OCR part and would appreciate your part of contribution. actually i am working on my thesis and i need to read document that consist of Mark and some text character. Your contribution would be very helpful.
Also is it possible to detect cross mark and remove them from image or replace them with white.. as this would be helpful to remove wrong mark on answer options.
Thank you

from omrchecker.

Hi @Adwait01-Bhosale, that's great to hear! Can you and @shreekrishnaacharya join on over discord and discuss possible solutions with me?

from omrchecker.

Here is a related issue for which I have framed a solution - #73

Just like extracting text from an area, that issue requires extracting passport photo from an area (using page cropping)

from omrchecker.

#99 (comment)
@Udayraj123 , @Adwait01-Bhosale my discord account is sksharma72000 . I have worked on tesseract OCR, it can detect printed text well however it is not suitable for handwriting detection.

from omrchecker.

That's fine for a start. I have a requirement from a business regarding the same. You can help read printed roll numbers or booklet numbers in an area on the image. Later we can scale it for handwriting detection

from omrchecker.

@shreekrishnaacharya @Adwait01-Bhosale let me know if you folks had any progress over this tasks. If so, others may decide to work on top of that.

from omrchecker.

from omrchecker.

@shreekrishnaacharya you can create a PR with a running code on one of the samples(make sure to take latest master pull)

p.s. or if you're available to discuss your approach first you can tag me on discord voice channel

from omrchecker.