bevaya/

SciTSR-pd

A public-domain subset of SciTSR, a large-scale table structure recognition dataset of scientific tables extracted from arXiv LaTeX source files. This subset contains only tables whose source papers carry a CC0 or equivalent public domain dedication — no attribution required, no restrictions on commercial or derivative use.

	Dataset Details




	

Split Tables Papers

train 89 —

test 19 —

total 108 52

Source paper licenses present: CC0… See the full description on the dataset page: https://huggingface.co/datasets/bevaya/SciTSR-pd.