why two different pre-trained models?

[mikequan0425]

I would like to ask why do you have two different pre-trained models for doctr and dewarpnet.Apart from training docproj datasets together, are there any differences between them?

[harrytea]

one is for the doctr geometric model and the other uses the dewarpnet geometric model.

[mikequan0425]

Are they all trained on docproj datasets? Since there are different models for the two geometric correction methods,I think they should have different settings at somewhere.