End-to-end text recognition with convolutional neural networks

By Tao Wang, David J. Wu, Adam Coates, Andrew Y. Ng. 2012

981 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Natural Language Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Natural Language Processing

Citations in the corpus

Deep Structured Output Learning for Unconstrained Text Recognition (2014, methodology): “Traditional text recognition methods are based on sequential character classiﬁcation, ﬁnding characters by sliding window methods (Wang et al. (2011; 2012); Jaderberg et al. (2014c), after which a word prediction is made by integrating character classiﬁer predictions in a left-to-right manner.”
Speeding up Convolutional Neural Networks with Low Rank Expansions (2014, background): “The training dataset consists of 163,222 collected character samples from a number of scene text and synthesized character datasets [1, 2, 5, 17, 21, 34, 42].”
Recent advances in convolutional neural networks (2015, background): “After several convolutional and pooling layers, there may be one or more fully-connected layers which aim to perform high level reasoning [8–10].”
Synthetic Data for Text Localisation in Natural Images (2016, background): “Current end-to-end text spotting (detection and recognition) methods [1, 20, 44] boost performance by combining detection with text recognition.”
Synthetic Data and Artificial Neural Networks for Natural Scene Text Recognition (2014, methodology): “Both [26] and [27] use a small fixed lexicon as a language model to constrain word recognition.”
Reading Text in the Wild with Convolutional Neural Networks (2014, methodology): “Following the success of some synthetic character datasets (de Campos et al. 2009; Wang et al. 2012), we create a synthetic word data generator, capable of emulating the distribution of scene text images.”
End-to-End Text Recognition with Hybrid HMM Maxout Models (2013, methodology): “The alternative of using a time-consuming but highly accurate classiﬁer as in (Wang et al., 2012) is not practical to make the end-to-end system work in real-time.”
Synthetic Data for Deep Learning (2019, neutral): “, [10, 169, 628, 686] and”
Large-Scale Learnable Graph Convolutional Networks (2018, background): “ional layers are trainable. The networks can automatically decide what kind of features to extract by learning the weights in these trainable filters, thereby avoiding hand-crafted feature extraction [29]. In many real-world applications, the data can be naturally represented as graphs, such as soci”
VISUAL ATTENTION (2014, positive): “Perhaps the most common approach to image-based classification of character sequences involves combining a sliding window detector with a character classifier (Wang et al., 2012; Jaderberg et al., 2014b).”
Neural Networks and Deep Learning (2018, neutral): “Convolutional neural networks for natural language processing are discussed in [78, 79, 102, 227, 240, 517].”
Decoupled Attention Network for Text Recognition (2019, methodology): “(Wang et al. 2012) used a CNN to recognize characters and adopt a non-maximum suppression to obtain the ﬁnal predictions.”
PhotoOCR: Reading Text in Uncontrolled Conditions (2013, neutral): “[26] 70.”
Arbitrary-Oriented Scene Text Detection via Rotation Proposals (2017, methodology): “To consider more precise styles of text, [10], [31] apply multiple scales and ratios to the sliding window methods.”
Mask TextSpotter: An End-to-End Trainable Neural Network for Spotting Text with Arbitrary Shapes (2018, background): “[72] attempt to detect characters by a sliding-window-based detector with CNNs and then recognize each character by a character classifier.”
Deep learning for smart fish farming: applications, opportunities and challenges (2020, background): “Fish heads have relatively fixed shapes and colors that can be used to track individual fish (Butail & Paley, 2011 ; Wang et al., 2012).”
Deep Tracking on the Move: Learning to Track the World from a Moving Vehicle using Recurrent Neural Networks (2016, background): “Recent advances in machine learning, particularly those of deep neural networks, have demonstrated the ability to capture complex structure in the real world, and have led to signiﬁcant improvements in numerous computer vision and natural language processing applications ([6], [7], [8]).”
Multimodal Neural Language Models (2014, methodology): “To speed up computation, we follow [20] [21] and choose to learn our convolutional networks on small feature maps learned using k-means as opposed to the original images.”
Text-Attentional Convolutional Neural Network for Scene Text Detection (2015, methodology): “a number of background components are classified incorrectly as text by the CNN used in [1] and [2], while some ambigu-”