Aligning benchmark datasets for table structure recognition

1 code implementation • 1 Mar 2023 • Brandon Smock, Rohith Pesala, Robin Abraham

However, even if a dataset's annotations are self-consistent, there may be significant inconsistency across datasets, which can harm the performance of models trained and evaluated on them.

GriTS: Grid table similarity metric for table structure recognition

1 code implementation • 23 Mar 2022 • Brandon Smock, Rohith Pesala, Robin Abraham

In this paper, we propose a new class of metric for table structure recognition (TSR) evaluation, called grid table similarity (GriTS).

PubTables-1M: Towards comprehensive table extraction from unstructured documents

2 code implementations • CVPR 2022 • Brandon Smock, Rohith Pesala, Robin Abraham

We demonstrate that these improvements lead to a significant increase in training performance and a more reliable estimate of model performance at evaluation for table structure recognition.

object-detection Object Detection +3