Filters








441 Hits in 6.2 sec

Enhance to Read Better: A Multi-Task Adversarial Network for Handwritten Document Image Enhancement [article]

Sana Khamekhem Jemni and Mohamed Ali Souibgui and Yousri Kessentini and Alicia Fornés
2021 arXiv   pre-print
To the best of our knowledge, this is the first work to use the text information while binarizing handwritten documents.  ...  Extensive experiments conducted on degraded Arabic and Latin handwritten documents demonstrate the usefulness of integrating the recognizer within the GAN architecture, which improves both the visual quality  ...  From each, we took the most used database for handwritten text line image recognition: KHATT [35] and IAM [36] , to add degradation.  ... 
arXiv:2105.12710v2 fatcat:oqri452ocjeqraj4g6t2yxjrpu

Offline Hand Written Urdu Word Spotting using Random Data Generation

Faiq Faizan Farooqui, Muhammad Hassan, Muhammad Shahzad Younis, Muhammad Kashif Siddhu
2020 IEEE Access  
For the word spotting process, Histogram of Oriented Gradients (HOG) features are extracted from ligature images and then used to train a Long Short-Term Memory (LSTM) network for the classification task  ...  When it comes to handwritten Urdu documents, variation among the same words of various writers is significant.  ...  We proposed the methodology of generating handwritten ligature image database by segregating the ligatures from handwritten document image lines.  ... 
doi:10.1109/access.2020.3010166 fatcat:ybwmfq7dfnecvkiz7czszbnd5y

Deep Learning for Historical Document Analysis and Recognition—A Survey

Francesco Lombardi, Simone Marinai
2020 Journal of Imaging  
Our study analyzes the papers published in the last few years on this topic from different perspectives: we first provide a pragmatic definition of historical documents from the point of view of the research  ...  Guided by these tasks, we go through the different input-output relations that are expected from the used deep learning approaches and therefore we accordingly describe the most used models.  ...  Siamese networks are used in Reference [12] to extract text lines by comparing document images patches and estimating their similarity.  ... 
doi:10.3390/jimaging6100110 pmid:34460551 pmcid:PMC8321201 fatcat:nevh2ctshzfwtey4girgjtaftq

Separating Chinese Character from Noisy Background Using GAN

Bin Huang, Jiaqi Lin, Jinming Liu, Jie Chen, Jiemin Zhang, Yendo Hu, Erkang Chen, Jingwen Yan, Philippe Fournier-Viger
2021 Wireless Communications and Mobile Computing  
Separating printed or handwritten characters from a noisy background is valuable for many applications including test paper autoscoring.  ...  This paper proposes a method for separating Chinese characters based on generative adversarial network (GAN).  ...  In the results, we show the visual effect of separating the handwritten part and the printed part, respectively, of four different common Chinese fonts.  ... 
doi:10.1155/2021/9922017 fatcat:gx4ofxf33rg2lgcun2bdlvmg3m

Deep Reader: Information extraction from Document images via relation extraction and Natural Language [article]

Vishwanath D, Rohit Rahul, Gunjan Sehgal, Swati, Arindam Chowdhury, Monika Sharma, Lovekesh Vig, Gautam Shroff, Ashwin Srinivasan
2018 arXiv   pre-print
In this paper, we propose a novel enterprise based end-to-end framework called DeepReader which facilitates information extraction from document images via identification of visual entities and populating  ...  However, extracting characters/text alone is often insufficient for relevant information extraction as documents also have a visual structure that is not captured by OCR.  ...  Processing Handwritten Text A major challenge in extracting useful information from scanned text documents is the recognition of handwritten text(HTR) in addition to printed text.  ... 
arXiv:1812.04377v2 fatcat:mtot4zi3lbe5xkmppiw3eeafru

Text and Style Conditioned GAN for the Generation of Offline-Handwriting Lines

Brian L. Davis, Bryan S. Morse, Brian L. Price, Chris Tensmeyer, Curtis Wigington, Rajiv Jain
2020 British Machine Vision Conference  
This paper presents a GAN for generating images of handwritten lines conditioned on arbitrary text and latent style vectors.  ...  After training, the encoder network can extract a style vector from an image, allowing images in a similar style to be generated, but with arbitrary text.  ...  In contrast, we produce higher quality handwritten line images from arbitrarily long text (instead of just words), and we can extract styles from existing images to generate similarly styled images.  ... 
dblp:conf/bmvc/DavisMPTWJ20 fatcat:3evmafsqmjfuxm3dqs5pucvapu

Illegible Text to Readable Text: An Image-to-Image Transformation using Conditional Sliced Wasserstein Adversarial Networks

Mostafa Karimi, Gopalkrishna Veni, Yen-Yun Yu
2020 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)  
Automatic text recognition from ancient handwritten record images is an important problem in the genealogy domain.  ...  We tackle this problem by developing a handwritten-to-machine-print conditional Generative Adversarial network (HW2MP-GAN) model that formulates handwritten recognition as a text-Image-to-text-Image translation  ...  Text-Image-to-Text-Image translation problem using HW2MP-GAN This section talks about the performance of the proposed HW2MP-GAN for solving the Text-Image-to-Text-Image translation problem.  ... 
doi:10.1109/cvprw50498.2020.00284 dblp:conf/cvpr/KarimiVY20 fatcat:gpx7wlchnre3hg4rzi6f6xxaeu

Editorial for special issue on "Advanced Topics in Document Analysis and Recognition"

Cheng-Lin Liu, Andreas Dengel, Rafael Dueire Lins
2019 International Journal on Document Analysis and Recognition  
In "Handwritten Arabic Text Recognition Using Multi-Stage Sub-Core Shape HMMs," Irfan Ahmad and Gernot Fink present a multi-stage HMM-based text recognition system for handwritten Arabic.  ...  The predictions are used as input for the second stage which performs a bottom-up clustering to build baselines of text lines.  ... 
doi:10.1007/s10032-019-00342-z fatcat:t4f7ruobrbf47hvudczsrg4r2q

CArDIS: A Swedish Historical Handwritten Character and Word Dataset

Amir Yavariabdi, Huseyin Kusetogullari, Turgay Celik, Shivani Thummanapally, Sakib Rijwan, Johan Hall
2022 IEEE Access  
The samples in CArDIS are collected from 64, 084 Swedish historical documents written by several anonymous priests between 1800 and 1900.  ...  This paper introduces a new publicly available image-based Swedish historical handwritten character and word dataset named Character Arkiv Digital Sweden (CArDIS) (https://cardisdataset.github. io/CARDIS  ...  In the handwriting document images, the text blocks and lines, as well as the transcriptions are annotated.  ... 
doi:10.1109/access.2022.3175197 fatcat:3vab2zq3srapzjto6n3xa7pwne

Synthesis in Style: Semantic Segmentation of Historical Documents using Synthetic Data [article]

Christian Bartz, Hendrik Raetz, Jona Otholt, Christoph Meinel, Haojin Yang
2022 arXiv   pre-print
Using our approach, we can extract the semantic information from the intermediate feature maps and use it to generate ground truth labels.  ...  To investigate if our synthetic dataset can be used to segment the text in historical documents, we use it to train multiple supervised segmentation models and evaluate their performance.  ...  Extracting these classes from a document is useful for two reasons: It may help archivists identify pages where potentially valuable, handwritten annotations are located.  ... 
arXiv:2107.06777v3 fatcat:jcfjammvirenvhri5pifvtmnce

Semi-supervised Feature Learning For Improving Writer Identification [article]

Shiming Chen, Yisong Wang, Chin-Teng Lin, Weiping Ding, Zehong Cao
2018 arXiv   pre-print
Data augmentation is usually used by supervised learning approaches for offline writer identification, but such approaches require extra training data and potentially lead to overfitting errors.  ...  Tang also used words segmented from handwritten documents as elements to permute the texts to generate a significant number of images, which were subsequently converted to form handwritten pages.  ...  Preprocessing: For the ICDAR2013 dataset, the handwritten documents are segmented into line images by a line segmentation method [42] , and then the line images are split up using a sliding window approach  ... 
arXiv:1807.05490v3 fatcat:dfxtdkumiffetbtpoye4jp35te

Text and Style Conditioned GAN for Generation of Offline Handwriting Lines [article]

Brian Davis, Chris Tensmeyer, Brian Price, Curtis Wigington, Bryan Morse, Rajiv Jain
2020 arXiv   pre-print
This paper presents a GAN for generating images of handwritten lines conditioned on arbitrary text and latent style vectors.  ...  After training, the encoder network can extract a style vector from an image, allowing images in a similar style to be generated, but with arbitrary text.  ...  In contrast, we produce higher quality handwritten line images from arbitrarily long text (instead of just words), and we can extract styles from existing images to generate similarly styled images.  ... 
arXiv:2009.00678v1 fatcat:4aj7ut4jpnau5oeiaqach6ydoa

Handwritten text generation and strikethrough characters augmentation [article]

Alex Shonenkov, Denis Karachev, Max Novopoltsev, Mark Potanin, Denis Dimitrov, Andrey Chertok
2021 arXiv   pre-print
We apply a novel augmentation that simulates strikethrough text (HandWritten Blots) and a handwritten text generation method based on printed text (StackMix), which proved to be very effective in HTR tasks  ...  Extensive experiments on ten handwritten text datasets show that HandWritten Blots augmentation and StackMix significantly improve the quality of HTR models  ...  We proposed weakly-supervised learning to extract characters boundaries from training images. As an example, we generated pages of texts from different sources.  ... 
arXiv:2112.07395v1 fatcat:5kwex6tsm5az5iz5jejhr4vvwq

Unsupervised Training Data Generation of Handwritten Formulas using Generative Adversarial Networks with Self-Attention [article]

Matthias Springstein and Eric Müller-Budack and Ralph Ewerth
2021 arXiv   pre-print
The recognition of handwritten mathematical expressions in images and video frames is a difficult and unsolved problem yet.  ...  In this paper, we introduce a system that creates a large set of synthesized training examples of mathematical expressions which are derived from LaTeX documents.  ...  Figure 5 : Sample images generated by the proposed GAN model after 90 000 iterations from the extracted Im2Latex-100k dataset [8] .  ... 
arXiv:2106.09432v1 fatcat:2j5wi7e2x5hyjmh3ygf2673hfi

Illegible Text to Readable Text: An Image-to-Image Transformation using Conditional Sliced Wasserstein Adversarial Networks [article]

Mostafa Karimi, Gopalkrishna Veni, Yen-Yun Yu
2019 arXiv   pre-print
Automatic text recognition from ancient handwritten record images is an important problem in the genealogy domain.  ...  We tackle this problem by developing a handwritten-to-machine-print conditional Generative Adversarial network (HW2MP-GAN) model that formulates handwritten recognition as a text-Image-to-text-Image translation  ...  Text-Image-to-Text-Image translation problem using HW2MP-GAN This section talks about the performance of the proposed HW2MP-GAN for solving the Text-Image-to-Text-Image translation problem.  ... 
arXiv:1910.05425v1 fatcat:k7vbel5dungdfn7qws7qbeuswy
« Previous Showing results 1 — 15 out of 441 results