Datasets:
There is no lines field in the json.
#3
by
wamreyaz
- opened
Hey all,
Thanks for the amazing work on this dataset. However, the json formats between the PDFA release and the IDL release is quite different.
Seems like you people have done the histogram based processing for the PDFA data but not the IDL data.
I am not sure if this is the intended release, as the dataset card for IDL explicity says that a line field should exist.
Best,
Wamiq
@wamreyaz the 'text' are the lines if no extra dict keys are present under 'text'
Hey @rwightman ,
Thanks, for the reply. I am processing these pdf not using chug.
I was merely asking if there is a plan to release the OCR-IDL dataset with the linebreaks as in PDFA?
Otherwise, I will run the heuristic based linebreak myself.
Best,
Wamiq