search

DCGM / pero-ocr

BSD 3-Clause "New" or "Revised" License
47 stars 19 forks source link

where can we find the pretrained models? #19

Closed jwijffels closed 4 years ago

jwijffels commented 4 years ago

are there checkpoints of models which can be downloaded available somewhere?

michal-hradis commented 4 years ago

We have released layout/textline model and OCR model for european printed documents. If you are interested in other models, let me know. The link is now in README.md:

General layout analysis (printed and handwritten) with european printed OCR specialized to czech newspapers can be downloaded here (https://www.fit.vut.cz/~ihradis/pero/pero_eu_cz_print_newspapers_2020-07-28.tar.gz). These models are compatible with the develop branch.

jwijffels commented 4 years ago

yes, I'm testing out a bit of handwritten text recognition models - would be interested to get that model as well, not only layout

michal-hradis commented 4 years ago

You can try a handwriting recognition model in our web application at pero-ocr.fit.vutbr.cz. The model is able to recognize some older european scripts and it is specialized for Czech. If your interest is deeper, feel free to contact me via my university email.

jwijffels commented 4 years ago

I've tried it already at the web application, set up the web application as well locally and noticed the models are not there, hence my question here. My interest is in training a handwritten text recognition model on 18th-19th century Dutch texts.