Current Status and Performance Analysis of Table Recognition in Document Images with Deep Neural Networks

The first phase of table recognition is to detect the tabular area in a\ndocument. Subsequently, the tabular structures are recognized in the second\nphase in order to extract information from the respective cells. Table\ndetection and structural recognition are pivotal problems in the domain of\ntable understanding. However, table analysis is a perplexing task due to the\ncolossal amount of diversity and asymmetry in tables. Therefore, it is an\nactive area of research in document image analysis. Recent advances in the\ncomputing capabilities of graphical processing units have enabled deep neural\nnetworks to outperform traditional state-of-the-art machine learning methods.\nTable understanding has substantially benefited from the recent breakthroughs\nin deep neural networks. However, there has not been a consolidated description\nof the deep learning methods for table detection and table structure\nrecognition. This review paper provides a thorough analysis of the modern\nmethodologies that utilize deep neural networks. This work provided a thorough\nunderstanding of the current state-of-the-art and related challenges of table\nunderstanding in document images. Furthermore, the leading datasets and their\nintricacies have been elaborated along with the quantitative results. Moreover,\na brief overview is given regarding the promising directions that can serve as\na guide to further improve table analysis in document images.\n