ocropus-rpred [add here possible options] <input files>

  -m MODEL, --model MODEL
                        line recognition model
  -p PAD, --pad PAD     extra blank padding to the left and right of text line
  -N, --nonormalize     don't normalize the textual output from the recognizer
  --llocs               output LSTM locations for characters
  --alocs               output aligned LSTM locations for characters
  --probabilities       output probabilities for each letter

  -e, --nolineest       deactivate the text normalizer and therefore the dewarping
  -l HEIGHT, --height HEIGHT
                        overrides the target line height, which is 48

  -r, --estrate         estimate error rate only
  -c ESTCONF, --estconf ESTCONF
                        estimate confusion matrix
  -C COMPARE, --compare COMPARE
                        string comparison used for error rate estimate
  --context CONTEXT     context for error reporting

  -s SHOW, --show SHOW  if >0, shows recognition output in a window and waits
                        this many seconds
  -S SAVE, --save SAVE  save debugging output image as PNG (for bug reporting)
  -h, --help            show this help message and exit
  -n, --nocheck         disable error checking on inputs
  -q, --quiet           turn off most output
  -Q PARALLEL, --parallel PARALLEL
                        number of parallel processes to use, default: 1