Recognition of Tables and Forms

Coüasnon, Bertrand; Lemaitre, Aurélie

doi:10.1007/978-0-85729-859-1_20

Cited by 52 publications

(23 citation statements)

References 46 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Hu et al [35] proposed a table detection method relying on the correlation of white spaces and vertical connected component analysis. For the comprehensive summarization of these rule-based approaches, readers may refer to [13,[36][37][38][39]. Although these rule-based methods work well on documents with similar tabular layouts, they are laborious in terms of finding optimal heuristics.…”

Section: Rule-based Approachesmentioning

confidence: 99%

See 1 more Smart Citation

HybridTabNet: Towards Better Table Detection in Scanned Document Images

Afzal¹,

Hashmi²,

Liwicki³

et al. 2021

Preprint

View full text Add to dashboard Cite

Tables in the document image are one of the most important entities since they contain crucial information. Therefore, accurate table detection can significantly improve information extraction from tables. In this work, we present a novel end-to-end trainable pipeline, HybridTabNet, for table detection in scanned document images. Our two-stage table detector uses the ResNeXt-101 backbone for feature extraction and Hybrid Task Cascade (HTC) to localize the tables in scanned document images. Moreover, we replace conventional convolutions with deformable convolutions in the backbone network. This enables our network to detect tables of arbitrary layouts precisely. We evaluate our approach comprehensively on ICDAR-13, ICDAR-17 POD, ICDAR-19, TableBank, Marmot, and UNLV. Apart from the ICDAR-17 POD dataset, our proposed HybridTabNet outperforms earlier state-of-the-art results without depending on pre and post-processing steps. Furthermore, to investigate how the proposed method generalizes unseen data, we conduct an exhaustive leave-one-out-evaluation. In comparison to prior state-of-the-art results, our method reduces the relative error by 27.57% on ICDAR-2019-TrackA-Modern, 42.64% on TableBank (Latex), 41.33% on TableBank (Word), 55.73% on TableBank (Latex + Word), 10% on Marmot, and 9.67% on UNLV dataset. The achieved results reflect the superior performance of the proposed method.

show abstract

Section: Rule-based Approachesmentioning

confidence: 99%

“…2) High intra-class variance (within the single class such as tables with and without ruling lines). Due to these chal-lenges, it is highly complex to come up with custom heuristics that can assist in developing robust and generic table detection system [13].…”

Section: Introductionmentioning

confidence: 99%

HybridTabNet: Towards Better Table Detection in Scanned Document Images

Afzal¹,

Hashmi²,

Liwicki³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Thus, a new table template matching that can deal with such variations has been developed. It matches the hierarchical structure of the table document and the defined template using an association graph (see Pelillo et al [1] and Ishitano [2]) by finding a maximum clique. Thus, the columns and the defined header in the template are detected.…”

Section: Template Processingmentioning

confidence: 99%

“…Among many other difficulties faced when designing a full Information Extraction workflow, we address in this paper the problem of table understanding: recognizing the structural organization of tables to extract data [1]. While in the considered documents (register books, see description Section 2), the vertical structures are relatively simple (nonhierarchical table composed of roughly 7-10 columns), the segmentation into rows turns out to be more challenging.…”

Section: Introductionmentioning

confidence: 99%

Comparing Machine Learning Approaches for Table Recognition in Historical Register Books

Clinchant¹,

Dejean²,

Meunier³

et al. 2018

2018 13th IAPR International Workshop on Document Analysis Systems (DAS)

View full text Add to dashboard Cite

We present in this paper experiments on Table Recognition in hand-written registry books. We first explain how the problem of row and column detection is modelled, and then compare two Machine Learning approaches (Conditional Random Field and Graph Convolutional Network) for detecting these table elements. Evaluation was conducted on death records provided by the Archive of the Diocese of Passau. Both methods show similar results, a 89 F1 score, a quality which allows for Information Extraction. Software and dataset are open source/data.

show abstract

“…Line segment recognition has, however, been steadily improved during the last three decades as part of table interpretation [1,2,3,4,5], form processing [6,7,8,9], and engineering drawing analysis [10,11,12,13]. Historical form analysis [14,15] became popular even as most contemporary forms migrated to the web. The Hough transform for line location has remained one of the leading methods for line and arc extraction since its rediscovery by Duda and Hart in the early seventies [16].…”

Section: Prior Workmentioning

confidence: 99%

On Parallel Lines in Noisy Forms

Nagy

2014

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. Quantification of the rectilinear configuration of typeset rules (lines) opens the way to form classification and content extraction. Line detection on scanned forms is often accomplished with the Hough transform. Here it is followed by simultaneous extraction of the dominant perpendicular sets of extracted lines, which ensures rotation invariance. Translation and scale invariance are attained by using minimal horizontal and vertical sets of distance ratios ("rule gap ratios") instead of rule-edge locations. The ratios are logarithmically mapped to an alphabet so that the resulting symbol strings can be classified by. edit distance. Some probability distributions associated with these steps are derived. Analytical considerations and small-scale experiments on scanned forms suggest that this approach has potential merit for highly degraded forms.

show abstract

Recognition of Tables and Forms

Cited by 52 publications

References 46 publications

HybridTabNet: Towards Better Table Detection in Scanned Document Images

HybridTabNet: Towards Better Table Detection in Scanned Document Images

Comparing Machine Learning Approaches for Table Recognition in Historical Register Books

On Parallel Lines in Noisy Forms

Contact Info

Product

Resources

About