When training from scratch, and especially for different languages, I would like to train a custom DocumentGenerator given my text dataset. It would be as simple as set an optional parameter gen=None to the function get_text_generator of data_generation. If it's None, then create it as it's being created now.
https://github.com/faustomorales/keras-ocr/blob/v0.8.5/keras_ocr/data_generation.py#L119
When training from scratch, and especially for different languages, I would like to train a custom
DocumentGenerator
given my text dataset. It would be as simple as set an optional parametergen=None
to the functionget_text_generator
ofdata_generation
. If it's None, then create it as it's being created now.