Li, Zhuoming; Peng, Fan; Xue, Yang; Ni, Hao; Jin, Lianwen; (2023) Scene Table Structure Recognition with Segmentation and Key Point Collaboration. In: Fink, Gernot A and Jain, Rajiv and Kise, Koichi and Zanibbi, Richard, (eds.) Document Analysis and Recognition - ICDAR 2023. (pp. pp. 295-310). Springer: Cham, Switzerland.

Preview

Text 2023_ICDAR_paper_2326.pdf - Accepted Version Download (2MB) | Preview

Abstract

This paper proposes a Segmentation and Key point Collaboration Network (SKCN) for structure recognition of complex tables with geometric deformations. First, we combine the cell regions of the segmentation branch and the corner locations of the key point regression branch in the SKCN to obtain more reliable detection bounding box candidates. Then, we propose a Centroid Filtering-based Non-Maximum Suppression algorithm (CF-NMS) to deal with the problem of overlapping detected bounding boxes. After obtaining the bounding boxes of all cells, we propose a post-processing method to predict the logical relationships of cells to finally recover the structure of the table. In addition, we design a module for online generation of tabular data by applying color, shading and geometric transformation to enrich the sample diversity of the existing natural scene table datasets. Experimental results show that our method achieves state-of-the-art performance on two public benchmarks, TAL_OCR_TABLE and WTW.

Download activity - last month

Export as JSONCSVXML

Download activity - last 12 months

Export as JSONCSVXML

Downloads by country - last 12 months

Export as CSVXMLJSON

Archive Staff Only

View Item