RNTR-Net: A Robust Natural Text Recognition Network
In this work, a novel robust natural text recognition network (RNTR-Net) is proposed based on a combination of convolutional neural network (CNN) (for feature extraction) and a recurrent neural network (RNN) (for sequence recognition). The pipeline design comprises an improved block of residual lear...
Main Authors: | , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
IEEE
2020-01-01
|
Series: | IEEE Access |
Subjects: | |
Online Access: | https://ieeexplore.ieee.org/document/8950043/ |
id |
doaj-647cd6203f8a469d99bdb3a63ad5f01a |
---|---|
record_format |
Article |
spelling |
doaj-647cd6203f8a469d99bdb3a63ad5f01a2021-03-30T01:18:37ZengIEEEIEEE Access2169-35362020-01-0187719773010.1109/ACCESS.2020.29641488950043RNTR-Net: A Robust Natural Text Recognition NetworkQiaokang Liang0https://orcid.org/0000-0002-5504-9966Shao Xiang1https://orcid.org/0000-0002-2797-1937Yaonan Wang2https://orcid.org/0000-0002-8578-2165Wei Sun3https://orcid.org/0000-0001-9744-4238Dan Zhang4https://orcid.org/0000-0002-7295-4837College of Electrical and Information Engineering, Hunan University, Changsha, ChinaCollege of Electrical and Information Engineering, Hunan University, Changsha, ChinaCollege of Electrical and Information Engineering, Hunan University, Changsha, ChinaCollege of Electrical and Information Engineering, Hunan University, Changsha, ChinaDepartment of Mechanical Engineering, York University, Toronto, ON, CanadaIn this work, a novel robust natural text recognition network (RNTR-Net) is proposed based on a combination of convolutional neural network (CNN) (for feature extraction) and a recurrent neural network (RNN) (for sequence recognition). The pipeline design comprises an improved block of residual learning combined with a general residual block to extract feature maps. Two bidirectional Long Short Term Memory (LSTM) networks are used for sequence recognition, and a transcription layer is used for decoding. The proposed network can handle text images suffering from distortion or other degradations. Compared with previous algorithms, we achieve superior results in general datasets, including the IIIT-5K, Street View Text and ICDAR datasets. Moreover, the performance of the presented network is either highly competitive or even state-of-the-art regarding the highly challenging SVT-Perspective and CUTE80 datasets. We obtain considerable performance of 84.7% and 62.6% on lexicon-free IIIT-5K and CUTE80 datasets, respectively. The experimental results demonstrate the effectiveness of our network.https://ieeexplore.ieee.org/document/8950043/Robust natural text recognition networkCNNresidual learningbidirectional LSTMs |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Qiaokang Liang Shao Xiang Yaonan Wang Wei Sun Dan Zhang |
spellingShingle |
Qiaokang Liang Shao Xiang Yaonan Wang Wei Sun Dan Zhang RNTR-Net: A Robust Natural Text Recognition Network IEEE Access Robust natural text recognition network CNN residual learning bidirectional LSTMs |
author_facet |
Qiaokang Liang Shao Xiang Yaonan Wang Wei Sun Dan Zhang |
author_sort |
Qiaokang Liang |
title |
RNTR-Net: A Robust Natural Text Recognition Network |
title_short |
RNTR-Net: A Robust Natural Text Recognition Network |
title_full |
RNTR-Net: A Robust Natural Text Recognition Network |
title_fullStr |
RNTR-Net: A Robust Natural Text Recognition Network |
title_full_unstemmed |
RNTR-Net: A Robust Natural Text Recognition Network |
title_sort |
rntr-net: a robust natural text recognition network |
publisher |
IEEE |
series |
IEEE Access |
issn |
2169-3536 |
publishDate |
2020-01-01 |
description |
In this work, a novel robust natural text recognition network (RNTR-Net) is proposed based on a combination of convolutional neural network (CNN) (for feature extraction) and a recurrent neural network (RNN) (for sequence recognition). The pipeline design comprises an improved block of residual learning combined with a general residual block to extract feature maps. Two bidirectional Long Short Term Memory (LSTM) networks are used for sequence recognition, and a transcription layer is used for decoding. The proposed network can handle text images suffering from distortion or other degradations. Compared with previous algorithms, we achieve superior results in general datasets, including the IIIT-5K, Street View Text and ICDAR datasets. Moreover, the performance of the presented network is either highly competitive or even state-of-the-art regarding the highly challenging SVT-Perspective and CUTE80 datasets. We obtain considerable performance of 84.7% and 62.6% on lexicon-free IIIT-5K and CUTE80 datasets, respectively. The experimental results demonstrate the effectiveness of our network. |
topic |
Robust natural text recognition network CNN residual learning bidirectional LSTMs |
url |
https://ieeexplore.ieee.org/document/8950043/ |
work_keys_str_mv |
AT qiaokangliang rntrnetarobustnaturaltextrecognitionnetwork AT shaoxiang rntrnetarobustnaturaltextrecognitionnetwork AT yaonanwang rntrnetarobustnaturaltextrecognitionnetwork AT weisun rntrnetarobustnaturaltextrecognitionnetwork AT danzhang rntrnetarobustnaturaltextrecognitionnetwork |
_version_ |
1724187251612581888 |