Can we use Tesseract 2.x trained files in Tesseract 3.02

Question

I have a set of trained files for "MICR" characters on Tesseract 2.04 platform. Is there any way that I can use the same trained data files in Tesseract 3.02 version?

I've tried to use "combine_tessdata" function on available data files, but failed. And I understand that there are major API changes going from 2.x version to 3.x version.

I'd be thankful to know whether I would have to train the characters again or there would exist method to use the current trained data files.

Thank you for your valuable time.

Answer 1

The language data between Tesseract 2.x and 3.x versions are not compatible. You would have to retrain them. The box files can be reused after adding a 6th, page number column.