I am trying to train Tesseract 4 with images instead of fonts. In docs, they only explain the approach in fonts, not images. I know how this works when I use the previous version of Tesseract, but I did not get how to use box / tiff files for training using LSTM in Tesseract 4. I looked at tesstrain.sh , which is used to generate LSTM , but did not find anything useful. Any ideas?
ocr tesseract training-data lstm
claim
source share