zilongwang/ReadingBank
ReadingBank ReadingBank is a benchmark dataset for reading order detection built with weak supervision from WORD documents, which contains 500K document images with a wide range of document types as well as the corresponding reading order information. Our paper "LayoutReader: Pre-training of Text and Layout for Reading Order Detection" has been accepted by EMNLP 2021. Refer to the official repo for more details: https://github.com/doc-analysis/ReadingBank
ReadingBank
ReadingBank is a benchmark dataset for reading order detection built with weak supervision from WORD documents, which contains 500K document images with a wide range of document types as well as the corresponding reading order information.
Our paper "LayoutReader: Pre-training of Text and Layout for Reading Order Detection" has been accepted by EMNLP 2021.
Refer to the official repo for more details: https://github.com/doc-analysis/ReadingBank