I am looking for some OCR engine (open source, preferably Python with Edge/Mobile devices support) that does not try to match found string (eg. CAPTCHA code but without image distortion like squeezing) to existing one in human dictionary. Eg. there is an image with text:
2ebra
or
15
I would NOT like to read it as:
Zebra
IS
Is there any OCR engine to support such exotic task? I am aware that some characters might be extremely similar, but I wish to proceed as little changes to found string as possible.
I have already tried to use Tesseract, but results are rather dissatisfactory.
I am looking forward to hearing it from you.