I have tested and reviewed the command options and don't see that there is a way to force gocr to make a quess, in other words, I am using the program to read a 5 digit invoice number. I have tried 200, 300, 400 dpi with no improvement over 200 dpi. The numbers are taken from white and yellow backgrounds but are converted, cleaned and dust removed. The program gets the number right about 50% of the time. Given the commnd:
gocr -d 20 -C "0123456789" -i test.jpg with only 10 digits to choose from I wish there was a option to force it to make the best quess instead of returning "123_56". I would think that if it was forced to guess it would probably get it right out of only the 10 choices. Any suggestions.
It is possible to do so, but its not a simple option (may be I should change that?).
Output in XML format. You will get a list of alternate chars. You need a XML-parser who makes the guess from that list and output as text.
May be you can send me a sample file and I can improve the engine and add a option to adjust the "accept-char" trigger value.
Is it already exist and what is option for use it with gocr?