Constraint output to HTML format
#73
by
MH1P
- opened
Hi all, and many thanks for the model!
I'm using the model to OCR some docs. It's pretty good at it, but I can't get it to generate an HTML output.
I'm trying to capture the structure of the document (so it goes beyond OCR) into HTML (e.g. using the proper title level, list, sublist, etc... no CSS) while ignoring headers and footers.
I have this "kind of working" with bigger commercial models, but can't have it working with idefics.
How to manage that? I'm trying to avoid finetuning at the moment as I don't have a 'Structured Aware OCR' dataset :-(
Cheers!