DocTr: Document Image Transformer for Geometric Unwarping and Illumination Correction

Feng, Hao; Wang, Yuechen; Zhou, Wengang; Deng, Jiajun; Li, Houqiang

doi:10.1145/3474085.3475388

Cited by 47 publications

(12 citation statements)

References 37 publications

(88 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In particular, Bako et al (2016) implemented a novel method based on explicitly identifying shadowed regions and increasing their brightness. Further, other approaches bridge multiple methods, such as Feng et al (2021), who implements a transformer machine learning model that simultaneously deskews scanned documents ("geometric unwarping") and removes their shadows ("illumination correction"). 4 .…”

Section: Color Corrections With Grayscale Outputmentioning

confidence: 99%

Digitizing Historical Balance Sheet Data: A Practitioner's Guide

Correia¹,

Luck²

2022

Preprint

View full text Add to dashboard Cite

This paper discusses how to successfully digitize large-scale historical micro-data by augmenting optical character recognition (OCR) engines with pre-and post-processing methods. Although OCR software has improved dramatically in recent years due to improvements in machine learning, off-the-shelf OCR applications still present high error rates which limits their applications for accurate extraction of structured information. Complementing OCR with additional methods can however dramatically increase its success rate, making it a powerful and cost-efficient tool for economic historians. This paper showcases these methods and explains why they are useful. We apply them against two large balance sheet datasets and introduce quipucamayoc, a Python package containing these methods in a unified framework.

show abstract

Section: Color Corrections With Grayscale Outputmentioning

confidence: 99%

Digitizing Historical Balance Sheet Data: A Practitioner's Guide

Correia¹,

Luck²

2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Recently, Amir et al [26] propose to learn the orientation of words in a document and Das et al [6] propose to model the 3D shape of a document with a UNet [32]. Feng et al [9] introduce transformer [36] from natural language processing tasks to improve the feature representation. Das et al [7] predict local deformation fields and stitch them together with global information to obtain an improved unwarping.…”

Section: Rectification Based On Deepmentioning

confidence: 99%

“…However, it involves extra implicit learning to localize the foreground document besides predicting the rectification, which limits the performance. Hence, following [9,10], we adopt a preprocessing operation to remove the clustered background first, thus the following network can focus on the rectification of the distortion.…”

Section: Preprocessingmentioning

confidence: 99%

“…Additionally, the 127 th and 128 th distorted images in DocUNet Benchmark dataset [25] are rotated by 180 degrees, which do not match the ground truth documents [10]. It is ignored by existing works [6,7,9,14,22,25,41,42,43]. In our experiments, we report the results on the corrected dataset.…”

Section: Revisiting Existing Datasetsmentioning

confidence: 99%

“…However, the oversimplified parametric models usually lead to limited performance, and the optimization process introduces non-negligible computational cost. Recently, deep learning based solutions [6,7,9,10,19,22,25] have been become a promising alternative to traditional methods. By training a network to directly predict the warping flow, a deformed document image can be rectified by resampling the pixels in the distorted image.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Geometric Representation Learning for Document Image Rectification

Feng¹,

Zhou²,

Deng³

et al. 2022

Preprint

View full text Add to dashboard Cite

In document image rectification, there exist rich geometric constraints between the distorted image and the ground truth one. However, such geometric constraints are largely ignored in existing advanced solutions, which limits the rectification performance. To this end, we present DocGeoNet for document image rectification by introducing explicit geometric representation. Technically, two typical attributes of the document image are involved in the proposed geometric representation learning, i.e., 3D shape and textlines. Our motivation raises from the insight that 3D shape provides global unwarping cues for rectifying a distorted document image, while overlooking the local structure. On the other hand, textlines complementarily provide explicit geometric constraints for local patterns. The learned geometric representation effectively bridges the distorted image and the ground truth one. Extensive experiments show the effectiveness of our framework and demonstrate the superiority of our DocGeoNet over state-of-the-art methods on both the DocUNet Benchmark dataset and our proposed DIR300 test set. Code is available at https://github.com/fh2019ustc/DocGeoNet.

show abstract