问题
I need to train tesseract to recognize just ten words.
Words are pharma name, thing like: Atrasil, Spectful
Since fonts used are pretty common, I tried to unpack eng.traineddata, substitute freq-dawg and word-dawg with just those words. Then I've repack them into a new traineddata, unfortunatly it doesn't seem to work very well. Matching results are still unacceptable and I can't use them, even when I use images obtained from a simple word file.
Is there a way to achieve a good matching? do I need to remove other files from the unpack trained data?
来源：https://stackoverflow.com/questions/25348214/tesseract-training-only-few-words