Deep Features for Text Spotting

By Max Jaderberg, Andrea Vedaldi, Andrew Zisserman. 2014

692 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Media Technology

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 Media Technology

Citations in the corpus

Deep Structured Output Learning for Unconstrained Text Recognition (2014, methodology): “Previous methods using CNNs for word recognition (discussed in more detail in section Section 2) has either constrained (Jaderberg et al. (2014b)) or heavily weighted (Bissacco et al. (2013)) the recognition results to be from a dictionary of known words.”
Speeding up Convolutional Neural Networks with Low Rank Expansions (2014, methodology): “Character classification is an essential part of many text spotting pipelines such as [3, 4, 15, 24, 25, 26, 27, 29, 30, 41, 43].”
Recent advances in convolutional neural networks (2015, methodology): “Going in a similar direction as [12], the CNN model proposed in [164] enables feature sharing across the four different subtasks of an end-to-end”
Synthetic Data for Text Localisation in Natural Images (2016, positive): “[44] and Jaderberg et al .”
Synthetic Data and Artificial Neural Networks for Natural Scene Text Recognition (2014, methodology): “These limitations have been mitigated before by mining for data or having access to large proprietary datasets [3, 11], but neither of these approaches are wholly accessible or scalable.”
Speeding-up Convolutional Neural Networks Using Fine-tuned CP-Decomposition (2014, methodology): “While (Jaderberg et al., 2014a) reported that such discriminative ﬁne-tuning was inefﬁcient for their scheme, we found that in our case it works well, even when CP-decomposition has large approximation error.”
Learning the Number of Neurons in Deep Networks (2016, methodology): “Therefore, similarly to [Jaderberg et al., 2014b, Denton et al., 2014] and to other parameter regularization techniques [Krogh and Hertz, 1992, Bartlett, 1996], these methods do not perform model selection; the number of layers and neurons per layer is determined manually and won’t be affected by…”
Reading Text in the Wild with Convolutional Neural Networks (2014, methodology): “Max Jaderberg · Karen Simonyan · Andrea Vedaldi · Andrew Zisserman Department of Engineering Science, University of Oxford E-mail: {max,karen,vedaldi,az}@robots.ox.ac.uk Text, as the physical incarnation of language, is one of the basic tools for preserving and communicating information.”
Total-Text: A Comprehensive Dataset for Scene Text Detection and Recognition (2017, methodology): “Alongside this work, [17], [18] also trained a CNN to classify text components from nontext.”
MaskOCR: Text Recognition with Masked Encoder-Decoder Pretraining (2022, background): “Text recognition can be achieved through various approaches, including character-based [22, 31, 34, 52], word-based [36], and sequence-based [12,44,45,57,60] methods.”
TextProposals: A text-specific selective search algorithm for word spotting in the wild (2016, methodology): “As far as end-to-end scene text recognition is concerned, most of the existing methods [14, 13, 33, 11, 25, 50, 28, 21, 12] rely in recognizing individual characters and then grouping them using spatial and linguistic constraints.”
Robust Seed Localization and Growing with Deep Convolutional Features for Scene Text Detection (2015, methodology): “[4] proposes a novel CNN architecture with eﬃcient feature sharing and other technical changes, which is used as the character classiﬁer and extended for character detection and bigram classiﬁcation.”
Read Like Humans: Autonomous, Bidirectional and Iterative Language Modeling for Scene Text Recognition (2021, background): “*The corresponding author methods [16, 14, 29] turn their attention to language modeling and achieve undoubted improvement.”
VISUAL ATTENTION (2014, neutral): “…multi-object and sequence recognition from images have avoided working with big images and instead focused on using ConvNets for recognizing characters or short sequence segments from image patches containing reasonably tightly cropped instances (Goodfellow et al., 2013; Jaderberg et al., 2014a).”
Towards End-to-End Text Spotting with Convolutional Recurrent Neural Networks (2017, methodology): “Previous works [28, 2, 12, 11] usually divide text spotting into a sequence of distinct sub-tasks.”
OCR-VQA: Visual Question Answering by Reading Text in Images (2019, background): “Especially, for the task of scene text recognition, we have witnessed significant boost in performance over the past years [5, 9, 12, 14, 25, 20].”
Symmetry-Constrained Rectification Network for Scene Text Recognition (2019, neutral): “ 87.6 - - - Almazan et al. [1] 91.2 82.1 - 89.2 - - - - - - - - Yao et al. [55] 80.2 69.3 - 75.9 - 88.5 80.3 - - - - - Rodr´ıguez-Serrano et al. [42] 76.1 57.4 - 70.0 - - - - - - - - Jaderberg et al. [22] - - - 86.1 - 96.2 91.5 - - - - - Su and Lu [47] - - - 83.0 - 92.0 82.0 - - - - - Gordo [12] 93.”
Automatic discrimination of text and non-text natural images (2015, background): “Note that, clustering is not applied to regions from background.”
Recognizing Text-Based Traffic Guide Panels with Cascaded Localization Network (2016, methodology): “[8]: a state-of-the-art method(5) that uses multiple stages of convolutional neural networks to predict text saliency score at each pixel, and cluster to form the region predictions afterward.”
Robust Scene Text Recognition with Automatic Rectification (2016, neutral): “[19] - - - 86.”