On cropped versus uncropped training sets in tabular structure detection

Abstract

Automated document processing for tabular information extraction is highly desired in many organizations, from industry to government. Prior works addressed this problem under table detection and table structure detection tasks. Proposed solutions that leverage deep learning approaches have been giving promising results; however, the impact of dataset structures on table structure detection has not been investigated. In this study, we provide a comparison of table structure detection performance with cropped and uncropped datasets. The cropped set consists of only table images that are cropped from documents assuming tables are detected perfectly. The uncropped set consists of regular document images. Experiments show that deep learning models can improve the detection performance by up to 9% in average precision and average recall on the cropped versions. Furthermore, the impact of cropped images is negligible under the Intersection over Union (IoU) values of 50%-70% when compared to the uncropped versions in small-scale datasets. However, beyond 70% IoU thresholds, cropped datasets provide significantly higher detection performance. In a large-scale dataset, cropped version outperforms the uncropped set under all IoU thresholds.