Hello, thanks for the project firstly, LSTM is working very well but i need to work in tesseract 2 for my university thesis.
I'm try training handwrite images (like 400 images) about 3 days and i cant successfull yet. Firstly i converted png images to tif and then i created box files for all of them (not with tesseract makebox because it is not recognizing characters in images) after that when i run box.train for images this problem appears. Why its happening ?
Also i was tried in tesseract 3 it was successful about creating TR files but after all traineddata is so bad for recognizing (300 Images Trained)
Thanks for all for responding