View Single Post
Old 25th May 2020, 14:01   #1018  |  Link
Janusz
Registered User
 
Join Date: Apr 2020
Location: Poland
Posts: 143
Quote:
Originally Posted by tormento View Post
What is nOCR?
Disable Subtitle Edit, in settings.ini find "<ShowBetaStuff>" and replace "False" with "True".
Launch the program. In [OCR Method] you will have a new method: "OCR via nOCR".
From the parameter name you can see that not everything can work as it should.
And that's how it is now. I didn't take notes of what I was doing and I can't reproduce what I wrote above. Fortunately, I have saved the character base and it can be repeated with it, but I can't generate the same database a second time.

---
For sure @Nikse555 will read it so I will add that:
the original character base entered from the hand to read the entire file error-free contains 367 elements, the new one was created by N-OCR training 481 characters so it may contain already recognized characters. I don't have the tools to check it.
---
It turned out that my admiration turned out to be premature. My mistake - I left my character base in the working directory, thanks to which the generated new characters were added to my base and hence the sensational result. Detriment. It seems that this project is no longer being developed. In the state in which it is now it can only serve as a curiosity.
--------------------------------------------------------------------------------------------

@Nikse555

There was a problem with beta 161.

- nOCR has stopped recognizing: . , - (three characters) and calls for each character encountered as a new one - unknown.

- 'o' recognizes as '0' or 'c', but this does not occur for everyone 'o' in the text.

Example:
beta 145: Nie. Na United Fusion Corporation. To co innego. (I worked on this version until then).
beta 161: Nie* Na United Fusion Corporation* T0 co inneg0*
__________________
Sorry for my mistakes - I'm using a translator.

Last edited by Janusz; 27th May 2020 at 07:26.
Janusz is offline   Reply With Quote