(Default) OCR in SDK, .dat files in Tesseract folder -> .traineddata SOLVED

[zarkogajic]

Hi Support,

Seems like in some of the few past releases, the extension of the language files for the Default OCR engine (the one available in SDK) has changed from ".dat" to ".traineddata".

Can you confirm that's the only change (related to using the default OCR from SDK)?

Are the same files used in both x86 and x64 (previously, with .dat, this was the case) ?


p.s.
For new readers, this is kind of an addon to this topic: viewtopic.php?t=33535

-žarko

[Tracker Supp-Stefan]

Hello zarkogajic,

I've passed your above enquiry to our devs working on the OCR engines - and we will post here a further update as soon as it's available!

Kind regards,
Stefan

[Vasyl-Tracker Dev Team]

Its occured because we updated the Tesseract engine. The new Tesseract uses the *.traineddata files instead of the older *.dat. And seems both formats are incompatible, unfortunately. And yes, both x86 and x64 use the same lang-files as well.

[zarkogajic]

Hi Vasyl,

Thanks

Btw,

And seems both formats are incompatible, unfortunately.

What do you mean?

I've simply renamed the file extension and all seems to work.

-žarko

[Tracker Supp-Stefan]

Hello zarkogajic,

I will ask Vasyl to clarify however if it works for you with just renaming the files - that's great!

Kind regards,
Stefan

[zarkogajic]

Ping

-žarko

[Vasyl-Tracker Dev Team]

I will ask my colleagues about your tricky method. Still not sure it is correct to just rename old lang files...

[Vasyl-Tracker Dev Team]

Our dev said that it was an upgrade of tesseract modules from v4 to newer v5. And newer tesseract uses different format for lang-files. Technically, the container is the same, but the data inside might be different. So it looks like the v5 tesseract is able to open and read those files, but there is a chance that some necessary data might be absent...

[zarkogajic]

Hi Vasyl,

Clear, thanks. Case closed.

-žarko

[Tracker Supp-Stefan]