Large-Lexicon Attribute-Consistent Text Recognition in Natural Images

Novikova, Tatiana; Barinova, Olga; Kohli, Pushmeet; Lempitsky, Victor

doi:10.1007/978-3-642-33783-3_54

Cited by 83 publications

(62 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Char detection Char classification High-level model Wang, ECCV '10 [26] Sliding window HOG + nearest-neighbor Pictorial structure Wang, ICCV'11 [27] Sliding window HOG + random ferns Pictorial structure Neumann, CVPR'12 [15] Extremal regions Shape + AdaBoost Pairwise rules Mishra, CVPR'12 [14] Sliding window HOG + SVM Pairwise CRF Mishra, BMVC'12 [13] Sliding window HOG + SVM High-order CRF Novikova, ECCV'12 [16] MSER HOG + nearest-neighbor Weighted finite-state transducer Table 1: Review of recent scene text recognition literature, analyzing the character detection, classification, and high-level model. HOG = histogram of gradients, SVM = support vector machine, CRF = conditional random field, MSER = maximally stable extrema region.…”

Section: Workmentioning

confidence: 99%

“…Despite building on the mature field of Optical Character Recognition (OCR), understanding text in natural scenes still poses significant challenges as indicated by recent papers [13,14,15,16,26,27]. Some difficulties are the use of multiple fonts, colors, or artistic designs, the textured backgrounds or the irregular character placement.…”

Section: Introduction and Related Workmentioning

confidence: 99%

“…The related works reviewed here [13,14,15,16,26,27] fit into this framework without exceptions, where the different components for character detection, classification, and high-level model are indicated in Table 1.…”

Section: Introduction and Related Workmentioning

confidence: 99%

See 2 more Smart Citations

Label embedding for text recognition

Rodríguez¹,

Perronnin²

2013

Procedings of the British Machine Vision Conference 2013

View full text Add to dashboard Cite

The standard approach to recognizing text in images consists in first classifying local image regions into candidate characters and then combining them with high-level word models such as conditional random fields (CRF). This paper explores a new paradigm that departs from this bottom-up view. We propose to embed word labels and word images into a common Euclidean space. Given a word image to be recognized, the text recognition problem is cast as one of retrieval: find the closest word label in this space. This common space is learned using the Structured SVM (SSVM) framework by enforcing matching label-image pairs to be closer than non-matching pairs. This method presents the following advantages: it does not require costly pre-or post-processing operations, it allows for the recognition of never-seen-before words and the recognition process is efficient. Experiments are performed on two challenging datasets (one of license plates and one of scene text) and show that the proposed method is competitive with standard bottom-up approaches to text recognition.

show abstract

Section: Workmentioning

confidence: 99%

Section: Introduction and Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Label embedding for text recognition

Rodríguez¹,

Perronnin²

2013

Procedings of the British Machine Vision Conference 2013

View full text Add to dashboard Cite

show abstract

“…One way to address the task of recognizing scene text is to pose the problem in conditional random field (crf) framework and obtain the maximum a posteriori (map) solution as proposed in [3,4,[7][8][9][10]. In these frameworks, an energy function consisting of unary and pairwise potentials is defined, and the minimum of this function corresponds to the text contained in the word image.…”

Section: Introductionmentioning

confidence: 99%

“…The problem of cropped word recognition has been looked at in two broad settings: with an image-specific lexicon [3][4][5][6]10] and without the help of lexicon [1,7,8]. Approaches for scene text recognition typically follow a two-step process (i) A set of potential character locations are detected either by binarization [1,2] or sliding windows [3,4], (ii) Inference on crf model [4,7], semi Markov model [1,8], finite automata [9] or beam search [2] in a graph (representing the character locations and their neighborhood relations) is performed. Fig.…”

Section: Introductionmentioning

confidence: 99%

Scene Text Recognition and Retrieval for Large Lexicons

Roy

Mishra

Alahari

et al. 2015

Computer Vision – ACCV 2014

View full text Add to dashboard Cite

Abstract. In this paper we propose a framework for recognition and retrieval tasks in the context of scene text images. In contrast to many of the recent works, we focus on the case where an image-specific list of words, known as the small lexicon setting, is unavailable. We present a conditional random field model defined on potential character locations and the interactions between them. Observing that the interaction potentials computed in the large lexicon setting are less effective than in the case of a small lexicon, we propose an iterative method, which alternates between finding the most likely solution and refining the interaction potentials. We evaluate our method on public datasets and show that it improves over baseline and state-of-the-art approaches. For example, we obtain nearly 15% improvement in recognition accuracy and precision for our retrieval task over baseline methods on the IIIT-5K word dataset, with a large lexicon containing 0.5 million words.

show abstract

Simultaneous Recognition of Horizontal and Vertical Text in Natural Images

Choi

Yoon

Lee

et al. 2019

Computer Vision – ACCV 2018 Workshops

View full text Add to dashboard Cite

Recent state-of-the-art scene text recognition methods have primarily focused on horizontal text in images. However, in several Asian countries, including China, large amounts of text in signs, books, and TV commercials are vertically directed. Because the horizontal and vertical texts exhibit different characteristics, developing an algorithm that can simultaneously recognize both types of text in real environments is necessary. To address this problem, we adopted the direction encoding mask (DEM) and selective attention network (SAN) methods based on supervised learning. DEM contains directional information to compensate in cases that lack text direction; therefore, our network is trained using this information to handle the vertical text. The SAN method is designed to work individually for both types of text. To train the network to recognize both types of text and to evaluate the effectiveness of the designed model, we prepared a new synthetic vertical text dataset and collected an actual vertical text dataset (VTD142) from the Web. Using these datasets, we proved that our proposed model can accurately recognize both vertical and horizontal text and can achieve state-of-the-art results in experiments using benchmark datasets, including the street view test (SVT), IIIT-5k, and ICDAR. Although our model is relatively simple as compared to its predecessors, it maintains the accuracy and is trained in an end-to-end manner.

show abstract

Large-Lexicon Attribute-Consistent Text Recognition in Natural Images

Cited by 83 publications

References 15 publications

Label embedding for text recognition

Label embedding for text recognition

Scene Text Recognition and Retrieval for Large Lexicons

Simultaneous Recognition of Horizontal and Vertical Text in Natural Images

Contact Info

Product

Resources

About