GitLab

traning with batch size = 1. Enable with -T flag
added resizing of codec

support for batch sizes

Added SequenceRecognizer dump into saving. Training on arbitrary batch size (None)

fixed loading of graph in tensorflow. rpred now supported

fixed codec for model loading

Added codec resizng (adding addition, removing unused)

Added codec resizng (adding addition, removing unused)

testing: move smoke tests in utils.py to run-unit

Native cleanup

Use case now handled by numpy in utils.py

Replace native code with regular functions

🐛 Do not add $datarootdir/ocropus to model if it is None

fix #267

Use numpy.einsum() for sumouter and sumprod functions.

minor typo fix

Fix regression in ocropus-gpageseg #251

Standardize Imports I, #176

# Conflicts:
#	ocropus-nlbin

Change checks for write_page_segmentation

Clip exponential in ffunc to avoid overflow

Ignored empty lines in fonts list file

Delete unused function bounding_box in ocropus-linegen

Fix ocropus-errs --erroronly

This function is now implemented in ocrolib/psegutils.py
and only used in ocropus-gpageseg not in the binarization
step.