Text Detection and Character Recognition in Scene Images with Unsupervised Feature Learning

By Adam Coates, Blake Carpenter, Carl Case, Sanjeev Satheesh, Bipin Suresh, Tao Wang, David J. Wu, Andrew Y. Ng. 2011

437 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Information Extraction

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 Information Extraction

Citations in the corpus

Deep Speech 2 : End-to-End Speech Recognition in English and Mandarin (2015, background): “Data augmentation has been highly effective in improving the performance of deep learning in computer vision (LeCun et al., 2004; Sapp et al., 2008; Coates et al., 2011) and speech recognition (Gales CTC”
Deep Speech: Scaling up end-to-end speech recognition (2014, methodology): “To expand our potential training data even further we use data synthesis, which has been successfully applied in other contexts to amplify the effective number of training samples [37, 26, 6].”
Deep learning for detecting robotic grasps (2013, neutral): “Only a handful of previous works have applied these methods to detection problems (LeCun et al., 2004; Osadchy et al., 2007; Coates et al., 2011).”
Learning with Hierarchical-Deep Models (2013, background): “Recently introduced deep learning models, including Deep Belief Networks [12], Deep Boltzmann Machines [29], deep autoencoders [19], and many others [9], [10], [21], [22], [2 6], [32], [34], [43], have been shown to learn useful distribute d feature representations for many high-dimensional datase t”
End-to-End Text Recognition with Hybrid HMM Maxout Models (2013, background): “Convolutional Neural Networks (CNNs) (LeCun et al., 1998) are discriminativly trained neural networks that alternate convolution layers and pooling layers, with the last layer being usually a softmax or RBF layer.”
TextProposals: A text-specific selective search algorithm for word spotting in the wild (2016, background): “[9] evaluate a single-layer Convolutional Neural Network model on each possible 32-by32 window of the input image at multiple scales.”
A Unified Framework for Multioriented Text Detection and Recognition (2014, methodology): “Due to its diversity, HUST-TR400 can serve as a standard benchmark for end-to-end scene text recognition.”
Multi-class Open Set Recognition Using Probability of Inclusion (2014, background): “Emerging research that moves beyond typical binary models of positive/negative class association for open set recognition has examined the issues of detecting novel classes [16,5,4], rejecting outlier or unknown classes [24,57,2], and/or simultaneously detecting and recognizing known classes in the ”
Text Detection and Recognition in Imagery: A Survey (2015, methodology): “vised [123] or representative learning [207], discriminative feature pooling [205], image rectification algorithms [132], or deformable models [195].”
Robust Scene Text Detection with Convolution Neural Network Induced MSER Trees (2014, methodology): “The structure of the CNN text classifier is similar to that applied in [29,4].”
Scene text detection and recognition: recent advances and future trends (2015, background): “[72] 2011 good performance on complex natural images robust only applicable to horizontal texts require a large volume of training data”
Deep Features for Text Spotting (2014, methodology): “Although the CNNs are large (2.6 million parameters for the case-insensitive classiﬁer), the method does not require any unsupervised pre-training as is done in [47], and incorporating the unsupervised approach described in [47, 13] gives no improvement.”
Unsupervised feature learning framework for no-reference image quality assessment (2012, background): “[3] showed the success of applying unsupervised feature learning in detecting text and recognizing characters in scene images.”
End-to-end text recognition with convolutional neural networks (2012, methodology): “We train the first layer of the network with an unsupervised learning algorithm similar to [2, 3].”
Recognizing Text with Perspective Distortion in Natural Scenes (2013, background): “The optimized alignment of the recognized characters with the lexicon will be discussed in the next section.”
Whole is Greater than Sum of Parts: Recognizing Scene Text Words (2013, background): “The document image analysis community has shown a huge interest in the problem of scene text understanding in recent years [6], [15], [19].”
Reading Digits in Natural Images with Unsupervised Feature Learning (2011, methodology): “We have also used the K-means-based feature learning system described by [18], which was recently applied to character recognition in [23].”