Filters








5,249 Hits in 4.2 sec

Random Blur Data Augmentation for Scene Text Recognition

Deguo Mu, Wei Sun, Guoliang Xu, Wei Li
2021 IEEE Access  
In this paper, we focus on Scene Text Recognition (STR) and propose two data augmentation approaches, named Random Blur Region (RBR) and Random Blur Units (RBUs).  ...  In summary, the main contributions of this paper are as follows: • We propose Random Blur Region (RBR), which is a simple and effective data augmentation approach for scene text recognition.  ... 
doi:10.1109/access.2021.3117035 fatcat:2yvlmrj6ubdshiahh2bdlx3rn4

Data Augmentation for Scene Text Recognition [article]

Rowel Atienza
2021 arXiv   pre-print
Scene text recognition (STR) is a challenging task in computer vision due to the large number of possible text appearances in natural scenes.  ...  The diversity and simplicity of API provided by STRAug functions enable easy replication and validation of existing data augmentation methods for STR.  ...  Thanks to CNL people: Roel Ocampo and Vladimir Zurbano, for hosting our servers.  ... 
arXiv:2108.06949v1 fatcat:7yvjar3aofhftitmselosmxgoi

Ultra Light OCR Competition Technical Report [article]

Shuhan Zhang, Yuxin Zou, Tianhe Wang, Yichao Xiong
2021 arXiv   pre-print
From experiments in aspects of data, model, training, etc, we proposed a general and effective method for Chinese scene text recognition, which got us second place among over 100 teams with accuracy 0.817  ...  Ultra Light OCR Competition is a Chinese scene text recognition competition jointly organized by CSIG (China Society of Image and Graphics) and Baidu, Inc.  ...  In natural scenes, text appears in various shapes and distorted patterns. In addition, the phenomenon of occlusion is also a big challenge for text recognition.  ... 
arXiv:2110.12623v1 fatcat:jvfowzo6lbh3rlkzful6yn3dze

A robust and effective text detector supervised by Contrastive Learning

Ran Wei, Yaoyi Li, Haiyan Li, Ze Tang, Hongtao Lu, Nengbin Cai, Xuejun Zhao
2021 IEEE Access  
INDEX TERMS Scene text detection, contrastive learning, data augmentation.  ...  So far, the detection results for text instances in motion blur, low-resolution images are still not satisfactory.  ...  TEXT DETECTION For a long period of time, scene text detection and recognition in natural scenes have been popular research topics in computer vision.  ... 
doi:10.1109/access.2021.3057108 fatcat:apcqjj76prh3bh3dfmr2a326n4

Scene Text recognition with Full Normalization [article]

Nathan Zachary, Gerald Carl, Russell Elijah, Hessi Roma, Robert Leer, James Amelia
2021 arXiv   pre-print
Scene text recognition has made significant progress in recent years and has become an important part of the work-flow.  ...  The widespread use of mobile devices opens up wide possibilities for using OCR technologies in everyday life. However, lack of training data for new research in this area remains relevant.  ...  of scene text detection and recognition research.  ... 
arXiv:2109.01034v1 fatcat:2zdpkyagqfbr3kgeriencv6etu

Learn to Augment: Joint Data Augmentation and Network Optimization for Text Recognition [article]

Canjie Luo, Yuanzhi Zhu, Lianwen Jin, Yongpan Wang
2020 arXiv   pre-print
In contrast to data collection and annotation, data augmentation is a low cost way. In this paper, we propose a new method for text image augmentation.  ...  Handwritten text and scene text suffer from various shapes and distorted patterns. Thus training a robust recognition model requires a large amount of data to cover diversity as much as possible.  ...  Irregular Scene Text Recognition Irregular shape is one of the challenges for scene text recognition. ASTER proposed by Shi et al.  ... 
arXiv:2003.06606v1 fatcat:q5mkvygkcfc3jeeoafpesynzdi

Learn to Augment: Joint Data Augmentation and Network Optimization for Text Recognition

Canjie Luo, Yuanzhi Zhu, Lianwen Jin, Yongpan Wang
2020 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)  
In contrast to data collection and annotation, data augmentation is a low cost way. In this paper, we propose a new method for text image augmentation.  ...  Handwritten text and scene text suffer from various shapes and distorted patterns. Thus training a robust recognition model requires a large amount of data to cover diversity as much as possible.  ...  In this paper, we propose a new data augmentation method for text recognition, which is designed for sequence-like characters [36] augmentation.  ... 
doi:10.1109/cvpr42600.2020.01376 dblp:conf/cvpr/LuoZJW20 fatcat:ivv3clfktvb4ji3py5fkcarrkq

Sequence-to-Sequence Contrastive Learning for Text Recognition [article]

Aviad Aberdam, Ron Litman, Shahar Tsiper, Oron Anschel, Ron Slossberg, Shai Mazor, R. Manmatha, Pietro Perona
2020 arXiv   pre-print
To yield effective visual representations for text recognition, we further suggest novel augmentation heuristics, different encoder architectures and custom projection heads.  ...  We propose a framework for sequence-to-sequence contrastive learning (SeqCLR) of visual representations, which we apply to text recognition.  ...  Text Recognition Background Several architectures have been proposed over the years for recognition of scene text [40, 12, 53] and handwritten text [58, 43] .  ... 
arXiv:2012.10873v1 fatcat:2i4yud7ayzfavmangejytjm3iq

ICDAR 2021 Competition on Scene Video Text Spotting [article]

Zhanzhan Cheng, Jing Lu, Baorui Zou, Shuigeng Zhou, Fei Wu
2021 arXiv   pre-print
Due to various environmental interferences like motion blur, spotting scene video text becomes very challenging.  ...  However, only a little effort has put to spotting scene video text, in contrast to massive studies of scene text spotting in static images.  ...  blur, random rotation, random crop, and random horizontal flip  ... 
arXiv:2107.11919v1 fatcat:kcaut7bsffholb7azc6whpeuee

IFR: Iterative Fusion Based Recognizer For Low Quality Scene Text Recognition [article]

Zhiwei Jia and Shugong Xu and Shiyi Mu and Yue Tao and Shan Cao and Zhiyong Chen
2021 arXiv   pre-print
In this paper, we propose an Iterative Fusion based Recognizer (IFR) for low quality scene text recognition, taking advantage of refined text images input and robust feature representation.  ...  IFR contains two branches which focus on scene text recognition and low quality scene text image recovery respectively.  ...  Owing to the good quality of the two synthetic datasets, we use random data augmentation, Gaussian kernel and down-up sampling, to generate paired training data.  ... 
arXiv:2108.06166v1 fatcat:d6gecramf5bkjlfkuxip7hcqyi

SynthTIGER: Synthetic Text Image GEneratoR Towards Better Text Recognition Models [article]

Moonbin Yim, Yoonsik Kim, Han-Cheol Cho, Sungrae Park
2021 arXiv   pre-print
For successful scene text recognition (STR) models, synthetic text image generators have alleviated the lack of annotated text images from the real world.  ...  data.  ...  MJ [5] ST [2] Ours OCR in the wild consists of two sub-tasks, scene text detection (STD) and scene text recognition (STR). They require similar but different training data.  ... 
arXiv:2107.09313v1 fatcat:u7cogiaf2jbf7m63bxoivvk6uu

Text-DIAE: Degradation Invariant Autoencoders for Text Recognition and Document Enhancement [article]

Mohamed Ali Souibgui, Sanket Biswas, Andres Mafla, Ali Furkan Biten, Alicia Fornés, Yousri Kessentini, Josep Lladós, Lluis Gomez, Dimosthenis Karatzas
2022 arXiv   pre-print
In this work, we propose Text-Degradation Invariant Auto Encoder (Text-DIAE) aimed to solve two tasks, text recognition (handwritten or scene-text) and document image enhancement.  ...  Finally, we demonstrate that our method surpasses the state-of-the-art significantly in existing supervised and self-supervised settings in handwritten and scene text recognition and document image enhancement  ...  Implementations details of Handwritten and Scene Text Recognition. The acronyms STR and HTR stands for Scene Text Recognition and Handwritten Text Recognition respectively. Config.  ... 
arXiv:2203.04814v3 fatcat:w7vbzopsxngu7njknd6a2n7shi

What If We Only Use Real Datasets for Scene Text Recognition? Toward Scene Text Recognition With Fewer Labels [article]

Jeonghun Baek, Yusuke Matsui, Kiyoharu Aizawa
2021 arXiv   pre-print
Scene text recognition (STR) task has a common practice: All state-of-the-art STR models are trained on large synthetic data.  ...  In contrast to this practice, training STR models only on fewer real labels (STR with fewer labels) is important when we have to train STR models without synthetic data: for handwritten or artistic texts  ...  The former is referred to as scene text detection (STD), and the latter as scene text recognition (STR).  ... 
arXiv:2103.04400v2 fatcat:nhheisp7jjcp7eh7hljautnm7q

An OCR Classifier for Republican Chinese Newspaper Text – Creation and Post-processing

Konstantin Henke, Matthias Arnold
2021 Zenodo  
Based on the hypothesis that pre-training on extensive amounts of suitably augmented character images will increase the OCR accuracy for evaluation on real-life character image data, we generate synthetic  ...  training data.  ...  1 Ren et al. (2016): A CNN Based Scene Chinese Text Recognition Algorithm With Synthetic Data Engine Character Image Generation: Algorithm for Randomized Augmentation a Extract glyph images from font.  ... 
doi:10.5281/zenodo.5770957 fatcat:3pa2c3uiqnhyvm5bjzxqmiqts4

An OCR Classifier for Republican Chinese Newspaper Text – Creation and Post-processing

Konstantin Henke, Matthias Arnold
2021 Zenodo  
Based on the hypothesis that pre-training on extensive amounts of suitably augmented character images will increase the OCR accuracy for evaluation on real-life character image data, we generate synthetic  ...  training data.  ...  1 Ren et al. (2016): A CNN Based Scene Chinese Text Recognition Algorithm With Synthetic Data Engine Character Image Generation: Algorithm for Randomized Augmentation a Extract glyph images from font.  ... 
doi:10.5281/zenodo.5770958 fatcat:p2sz3ajbnvc5fit6brae7556au
« Previous Showing results 1 — 15 out of 5,249 results