Pith. sign in

REVIEW 3 major objections 5 minor 60 references

Symmetry-constrained Rectification Network for Scene Text Recognition

T0 review · 3 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

Pith's one-line read A symmetry-constrained rectification network predicts a text line's center, scale, and orientation, generates mirror-symmetric control points, and warps irregular scene text to a regular form, reporting top accuracy on curved benchmarks.

desk verdict Solid, well-controlled paper on geometry-based rectification for scene text recognition; the gains are real but the abstract oversells them and the terminal-character failure mode is a contained limitation. read the letter →

arxiv 1908.01957 v1 pith:QASDWTQY submitted 2019-08-06 cs.CV

classification cs.CV
keywords scenetextrecognitionirregularrectificationthin-platesplinecenterlinegeometricattributesspatialtransformernetworkattentiondecodersynthetictrainingdata
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper argues that rectifying irregular scene text before recognition fails for highly curved words because prior methods predict control points on the text outline independently, with no guarantee of symmetry about the text's center line. It proposes a Symmetry-constrained Rectification Network (ScRN), a two-convolution-layer module that predicts the text center line, per-pixel scale, text orientation, and character orientation, then derives control points that are exactly symmetric about the center line. These control points feed a thin-plate-spline warp that straightens the feature maps before an attention-based decoder reads them. If correct, the approach makes rectification more robust and interpretable, and the paper reports accuracy gains of roughly 1 to 8 percentage points over prior methods on irregular-text benchmarks such as ICDAR 2015, SVT-Perspective, and CUTE80, at negligible added inference cost.

What carries the argument

The central object is the text center line (TCL) with per-point geometric attributes $\mathrm{geo}_i = (c_i; s_i; \phi_i; \theta_i)$, where $c_i$ is a center point, $s_i$ is half the character height, $\phi_i$ is the character orientation, and $\theta_i$ is the text orientation tangent. Control points are generated by taking each of $k$ evenly spaced samples $\bar{c}_i$ and placing two points at distance $s_i$ along the character-orientation direction: $p_{2i-1} = \bar{c}_i + (s_i \cos\phi_i, -s_i \sin\phi_i)$ and $p_{2i} = \bar{c}_i - (s_i \cos\phi_i, -s_i \sin\phi_i)$, so the paired points are mirror-symmetric about the center line by construction. These points are then used as fiducials for a thin-plate-spline transformation that warps the shared feature maps to a canonical horizontal strip before an attention-based sequence decoder reads them. The character orientation $\phi$ is the distinctive addition: it lets the warp follow slanted or sheared characters rather than the normal to the center line.

What would settle it

Take a set of curved words whose end characters are nearly horizontal and close to the image border, measure the distance between ScRN's predicted control points and the true text edges, and compare recognition accuracy to the no-rectification baseline on that subset; if the control-point error is no smaller than the STN baseline's and the accuracy gain disappears, the central claim that symmetry-constrained center-line attributes improve rectification is false in the regime the paper identifies as its limitation.

Watch

Extended reading notes

Core claim

The central claim is that representing a text instance by its center line plus local geometrical attributes—scale, text orientation, and character orientation—and using those attributes to generate paired, mirror-symmetric control points for a thin-plate-spline warp yields better rectification than weakly supervised spatial-transformer networks that predict outline control points directly. Because the control points are constructed as equidistant samples along the predicted center line, offset by the predicted scale along the predicted character orientation, the upper and lower points are symmetric by construction. The paper reports that this symmetric construction, trained with explicit geometry supervision on synthetic data, outperforms its own STN-based baselines on IC15, SVTP, and CUTE80 by 0.5 to 1.7 percentage points and matches or exceeds them on regular-text benchmarks, with an additional 1 millisecond per image at inference.

Load-bearing premise

The whole method assumes that a word's shape is faithfully captured by a center line through character-box centers, with scale and orientation values linearly interpolated between centers; when terminal characters are nearly horizontal and touch the image border, this assumption breaks and the paper's own failure analysis shows misaligned control points.

Editorial extensions

If this is right

  • The rectification module itself, not the extra geometry loss, drives the gains: a multi-loss baseline without rectification matches the base recognizer, while the full ScRN improves irregular-text accuracy.
  • The added cost is small: prediction uses two convolutional layers and inference rises from 12 ms to 13 ms per image, so the module can be added to existing recognizers almost for free.
  • The character-orientation attribute is necessary: replacing it with the center-line normal degrades rectification visibly, so the paper's representation is load-bearing rather than cosmetic.
  • Because geometry supervision comes from SynthText character boxes, word-level-only corpora such as Synth90k can still be added for recognition training without needing extra geometry labels.
  • On irregular-text datasets the reported gains over the STN baseline are 0.5 points on IC15, 1.4 points on SVTP, and 1.7 points on CUTE80, and the gains over the weakly supervised STN with extra supervision are consistent but smaller.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A natural extension the authors leave implicit is to unify detection and recognition through the same center-line representation, allowing end-to-end recognition of arbitrary-shaped words without separate cropping.
  • A testable prediction is that rectification quality should degrade as the predicted center line is corrupted; injecting controlled noise into $\phi$ and $s$ would isolate how much of the gain is due to the symmetry constraint and how much to the geometry itself.
  • The acknowledged endpoint failure suggests a concrete fix: extend the representation at the line ends, for example with per-character quadrilaterals, to remove linear-interpolation error where terminal characters are nearly horizontal.
  • An unstated consequence of the shared-feature design is that rectifying feature maps rather than the input image is what keeps the module cheap; applying ScRN to full-resolution images would let existing recognizers benefit at roughly twice the compute.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes a Symmetry-constrained Rectification Network (ScRN) for scene text recognition. ScRN predicts per-pixel text center lines, scale, text orientation, and character orientation from shared FPN features, generates 2k Thin-Plate-Spline control points symmetrically around the center line via Eq. (2), and rectifies the shared feature maps before an attention-based recognizer. The model is trained end-to-end with a geometry loss and a recognition loss on SynthText and Synth90k, and evaluated on seven standard benchmarks. The central claims are that ScRN achieves state-of-the-art accuracy on regular and irregular text, outperforming existing methods by a large margin on ICDAR 2015, SVT-Perspective, and CUTE80, while adding negligible computation through a two-convolution-layer rectification module.

Significance. If the reported results are reproducible, ScRN is a useful and practical contribution: it gives an interpretable, symmetry-constrained alternative to STN-based rectification, with controlled experiments that correctly isolate the rectification module from the additional geometry loss. The paper's controlled comparisons in Tables 2 and 3 are a genuine strength and rule out the common confound that the gains come only from the extra loss. The computational overhead is small (about 1 ms per image), and the visualizations in Figs. 6 and 7 support the claimed qualitative behavior. However, the significance is currently limited by three issues: the controlled gains over a same-backbone supervised STN baseline are modest (0.2-1.7 points), the results appear to be single-run with no error bars or released code, and the method's acknowledged endpoint failure in Sec. 4.6 is not quantified despite being directly relevant to the headline irregular-text claim.

major comments (3)
  1. [Sec. 3.1.3, Eq. (2), and Sec. 4.6] The failure mode reported in Sec. 4.6 is a direct consequence of the control-point parameterization in Eq. (2). When a terminal character's orientation phi_i is nearly horizontal, the two control points p_{2i-1} and p_{2i} lie almost on the center line rather than on opposite text borders, so the TPS envelope is locally unconstrained at the word boundary. Because the abstract's headline result is the large margin on the irregular benchmarks, where curved text often has near-horizontal terminal characters close to image borders, this acknowledged limitation needs quantitative treatment. Please report the frequency of such cases on IC15, SVTP, and CUTE80, show example-level rectification quality at word endpoints, and state whether a simple post-processing constraint (e.g., clamping phi near the borders) would remove the failure.
  2. [Sec. 4.5/Table 4 vs Sec. 4.4/Table 3] The 'large margin' claim over ASTER on IC15, SVTP, and CUTE80 in the abstract and Sec. 4.5 is based on cross-method comparison with different training data and system components. In the controlled same-backbone, same-training comparison of Table 3, ScRN exceeds the STN_supervision baseline by only 0.2, 1.1, and 1.0 points on IC15, SVTP, and CUTE80, respectively. The paper should either calibrate the abstract and conclusion to the controlled comparison or provide evidence that the Table 4 differences are attributable to the rectification module rather than to other differences between the systems.
  3. [Sec. 4.2/4.4] All results appear to be from a single run and no code is provided. Since the controlled gains in Table 3 are as small as 0.2-1.7 points, the paper should report the number of runs and variance (or bootstrap confidence intervals) for at least the main comparisons in Tables 2 and 3, and should release the code or a sufficiently detailed protocol to make the experiments reproducible.
minor comments (5)
  1. [Sec. 3.1.2, Eq. (1)] Equation (1) normalizes cos(phi) and sin(phi), but the text says cos(theta) and sin(theta) are normalized in the same way without giving the analogous equation; please add it for completeness.
  2. [Table 4] The column headers of Table 4 ('50 1k 0 50 0 50 Full 0 0 0 0 0') are hard to parse; spell out the evaluation protocol for each dataset, including which columns correspond to lexicon-free recognition.
  3. [Sec. 4.4] The STN baseline and STN supervision are said to share the same backbone and recognition module, but the STN architecture details (number of layers, control-point count, training details) are not fully specified beyond 'similar with ASTER'; include enough details to reproduce the ablation.
  4. [Sec. 3.3.1] Equation (4) defines Lgeo only for SynthText; clarify that Synth90k examples contribute only the recognition loss in both training stages, and how the two-stage protocol treats un-annotated geometry in the second stage.
  5. [Sec. 4.6] The phrase 'labor-free and time-efficient' is misleading because character-level annotations are still required; rephrase to emphasize that the annotations are obtained automatically from the synthesizing engine rather than by manual labeling.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the SOTA claim is an empirical benchmark result, and the rectification geometry is defined and trained independently of those benchmarks.

full rationale

The paper's central claim is empirical recognition accuracy on standard external benchmarks (IIIT5K, SVT, IC03, IC13, IC15, SVTP, CUTE80). The rectification module predicts center line, scale, and orientation attributes from a two-convolution-layer network supervised by ground-truth geometric attributes on SynthText (Sec. 3.3, Eq. 5), then generates TPS control points via Eq. (2). None of the benchmark accuracies are used to define, fit, or post-hoc select the geometric predictor; the recognition loss (Eq. 3) and geometry loss (Eq. 5) are conventional supervised losses. The center-line/scale/orientation representation is explicitly adopted from prior work (TextSnake, [33]) and compared against ASTER ([46]), with which some authors overlap, but the paper itself defines the representation in Sec. 3.1.1 and the control-point generation in Sec. 3.1.3, and the cited works have independent empirical content and are not invoked as a uniqueness theorem or to forbid alternatives. The acknowledged limitation in Sec. 4.6, where terminal characters that are nearly horizontal and close to image borders cause imperfect rectification, is an empirical failure mode, not a circular reduction. No equation or fitted parameter in the derivation chain is equivalent by construction to the claimed prediction, so there is no circularity to report.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central claim depends on hand-set hyperparameters (loss weights, number of control points, TCL expansion) and on the chosen text-geometry representation. No mathematical derivation or fitted theory is involved; the contribution is empirical architecture design.

free parameters (3)
  • Loss weights lambda_1 through lambda_6 = 1 (all)
    Eqn. (5) and Sec. 3.3.1 set all six geometry-loss weights to 1 with no ablation; the balance between geometry and recognition losses is a design choice that affects the trained rectifier.
  • Number of sampled control points k = 10
    Sec. 4.2 fixes k=10 for the center-line resampling to build mini-batches; this controls TPS flexibility and is chosen by hand.
  • TCL expansion radius = 1 pixel
    Sec. 4.2 expands the text center line by one pixel to reduce noise; this expansion affects the supervision mask for geometry prediction.
assumptions (5)
  • domain assumption A text instance can be represented by a center line built from character-box centers, with per-pixel scale, character orientation, and text orientation linearly interpolated between centers.
    Sec. 3.1.1 defines the representation and interpolation; the rectification control points are derived entirely from this representation, so if the interpolation model fails on real curved text, the rectification degrades (failure cases in Sec. 4.6).
  • domain assumption Geometric attributes predicted by a two-convolution-layer network trained only on SynthText generalize to real-world irregular text.
    No real images with character-level geometry annotations are used for geometry supervision (Sec. 3.3 and Sec. 4.2); the claimed accuracy on real benchmarks depends on this transfer.
  • domain assumption Character orientation is a necessary attribute in addition to text orientation for accurate rectification.
    Sec. 3.1.4 and Fig. 5 argue this with an example but do not provide a quantitative ablation; if character orientation is redundant or poorly predicted, the added complexity is not justified.
  • domain assumption Thin-plate spline warping with a fixed set of symmetric control points (k=10) can map irregular text feature maps to a canonical horizontal form.
    Sec. 3.1.3 and Sec. 4.2; final recognition accuracy depends on this warping being adequate for highly curved text.
  • standard math Thin-plate spline interpolation is differentiable and suitable for end-to-end training.
    Standard result cited from Bookstein [6] and used in Sec. 3.1.3.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Symmetry-constrained Rectification Network for Scene Text Recognition." pith.science (2026). https://pith.science/paper/QASDWTQY

@misc{pith2026190801957,
  author       = {Pith},
  title        = {Pith review of: Symmetry-constrained Rectification Network for Scene Text Recognition},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QASDWTQY}},
  note         = {Machine review of arXiv:1908.01957}
}
read the original abstract

Reading text in the wild is a very challenging task due to the diversity of text instances and the complexity of natural scenes. Recently, the community has paid increasing attention to the problem of recognizing text instances with irregular shapes. One intuitive and effective way to handle this problem is to rectify irregular text to a canonical form before recognition. However, these methods might struggle when dealing with highly curved or distorted text instances. To tackle this issue, we propose in this paper a Symmetry-constrained Rectification Network (ScRN) based on local attributes of text instances, such as center line, scale and orientation. Such constraints with an accurate description of text shape enable ScRN to generate better rectification results than existing methods and thus lead to higher recognition accuracy. Our method achieves state-of-the-art performance on text with both regular and irregular shapes. Specifically, the system outperforms existing algorithms by a large margin on datasets that contain quite a proportion of irregular text instances, e.g., ICDAR 2015, SVT-Perspective and CUTE80.

Figures

Figures reproduced from arXiv: 1908.01957 by the authors.

Figure 2
Figure 2. Pipeline of the proposed method. symmetrical constraints in their spatial distribution. ScRN is a simple segmentation network which only consists of two convolutional layers. Therefore, it just incurs negligi￾ble computation and storage overhead when combined with a text recognizer. Compared with the previous STN-based rectification methods, ScRN has superiorities in both ro￾bustness and interpretability, profiting … view at source ↗
Figure 3
Figure 3. Illustration of the text representation. [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figure 4
Figure 4. The rectification process. Note that, for all figures in this paper, we use the input image to illustrate these points [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figures from the paper (4 more)
Figure 5
Figure 5. Figure 5: Control points and rectification results using the [PITH_FULL_IMAGE:figures/full_fig_p004_5.png]
Figure 6
Figure 6. Figure 6: Selected results from SVTP and CUTE80, which suffer from severe distortion. For every three rows, the first row [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 7
Figure 7. Figure 7: Rectified results produced by our proposed ScRN, [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 8
Figure 8. Figure 8: Some bad cases produced by our recognition sys [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

60 extracted references · 56 canonical work pages

  1. [1]

    Almaz ´an, A

    J. Almaz ´an, A. Gordo, A. Forn ´es, and E. Valveny. Word spotting and recognition with embedded attributes. TPAMI, 36(12):2552–2566, 2014. 8

  2. [2]

    Bahdanau, K

    D. Bahdanau, K. Cho, and Y . Bengio. Neural machine translation by jointly learning to align and translate. CoRR, abs/1409.0473, 2014. 4

  3. [3]

    F. Bai, Z. Cheng, Y . Niu, S. Pu, and S. Zhou. Edit probability for scene text recognition. In CVPR, 2018. 2, 8

  4. [4]

    X. Bai, C. Yao, and W. Liu. Strokelets: A learned multi-scale mid-level representation for scene text recognition. IEEE Transactions on Image Processing, 25(6):2789–2802, 2016. 2

  5. [5]

    Bissacco, M

    A. Bissacco, M. Cummins, Y . Netzer, and H. Neven. Pho- toocr: Reading text in uncontrolled conditions. In ICCV, pages 785–792, 2013. 8

  6. [6]

    F. L. Bookstein. Principal warps: Thin-plate splines and the decomposition of deformations. TPAMI, 11(6):567–585,

  7. [7]

    Cheng, F

    Z. Cheng, F. Bai, Y . Xu, G. Zheng, S. Pu, and S. Zhou. Fo- cusing attention: Towards accurate text recognition in natural images. In ICCV, pages 5086–5094, 2017. 2, 8

  8. [8]

    Cheng, Y

    Z. Cheng, Y . Xu, F. Bai, Y . Niu, S. Pu, and S. Zhou. Aon: To- wards arbitrarily-oriented text recognition. In CVPR, pages 5571–5579, 2018. 1, 2, 7, 8

Show all 60 references
  1. [9]

    K. Cho, B. van Merrienboer, C ¸ . G ¨ulc ¸ehre, D. Bahdanau, F. Bougares, H. Schwenk, and Y . Bengio. Learning phrase representations using RNN encoder-decoder for statistical machine translation. In Proceedings of the 2014 Confer- ence on Empirical Methods in Natural Language...

  2. [10]

    Epshtein, E

    B. Epshtein, E. Ofek, and Y . Wexler. Detecting text in natural scenes with stroke width transform. In CVPR, pages 2963– 2970, 2010. 2

  3. [11]

    Girshick

    R. Girshick. Fast r-cnn. In ICCV, December 2015. 5

  4. [12]

    A. Gordo. Supervised mid-level features for word image rep- resentation. In CVPR, pages 2956–2964, 2015. 8

  5. [13]

    Graves, S

    A. Graves, S. Fern ´andez, F. J. Gomez, and J. Schmidhu- ber. Connectionist temporal classification: labelling unseg- mented sequence data with recurrent neural networks. In ICML, pages 369–376, 2006. 2

  6. [14]

    Graves, M

    A. Graves, M. Liwicki, S. Fern ´andez, R. Bertolami, H. Bunke, and J. Schmidhuber. A novel connectionist sys- tem for unconstrained handwriting recognition. TPAMI, 31(5):855–868, 2009. 4

  7. [15]

    Gupta, A

    A. Gupta, A. Vedaldi, and A. Zisserman. Synthetic data for text localisation in natural images. In CVPR, pages 2315– 2324, 2016. 5, 8

  8. [16]

    K. He, X. Zhang, S. Ren, and J. Sun. Deep residual learning for image recognition. In CVPR, pages 770–778, 2016. 2

  9. [17]

    P. He, W. Huang, Y . Qiao, C. C. Loy, and X. Tang. Reading scene text in deep convolutional sequences. In AAAI, vol- ume 16, pages 3501–3508, 2016. 2

  10. [18]

    Jaderberg, K

    M. Jaderberg, K. Simonyan, A. Vedaldi, and A. Zisserman. Synthetic data and artificial neural networks for natural scene text recognition. CoRR, abs/1406.2227, 2014. 5

  11. [19]

    Jaderberg, K

    M. Jaderberg, K. Simonyan, A. Vedaldi, and A. Zisserman. Deep structured output learning for unconstrained text recog- nition. In ICLR, 2015. 8

  12. [20]

    Jaderberg, K

    M. Jaderberg, K. Simonyan, A. Vedaldi, and A. Zisserman. Reading text in the wild with convolutional neural networks. IJCV, 116(1):1–20, 2016. 2, 8

  13. [21]

    Jaderberg, K

    M. Jaderberg, K. Simonyan, A. Zisserman, et al. Spatial transformer networks. In NIPS, pages 2017–2025, 2015. 1, 2

  14. [22]

    Jaderberg, A

    M. Jaderberg, A. Vedaldi, and A. Zisserman. Deep features for text spotting. In ECCV, pages 512–528, 2014. 8

  15. [23]

    Karatzas, L

    D. Karatzas, L. Gomez-Bigorda, A. Nicolaou, S. K. Ghosh, A. D. Bagdanov, M. Iwamura, J. Matas, L. Neumann, V . R. Chandrasekhar, S. Lu, F. Shafait, S. Uchida, and E. Valveny. ICDAR 2015 competition on robust reading. In Proc. IC- DAR, pages 1156–1160, 2015. 5

  16. [24]

    Karatzas, F

    D. Karatzas, F. Shafait, S. Uchida, M. Iwamura, L. G. i Big- orda, S. R. Mestre, J. Mas, D. F. Mota, J. A. Almazan, and L. P. de las Heras. Icdar 2013 robust reading competition. In ICDAR, pages 1484–1493, 2013. 5

  17. [25]

    Lee and S

    C.-Y . Lee and S. Osindero. Recursive recurrent nets with attention modeling for ocr in the wild. InCVPR, pages 2231– 2239, 2016. 2, 8

  18. [26]

    G. Li, S. Xu, X. Liu, L. Li, and C. Wang. Jersey num- ber recognition with semi-supervised spatial transformer net- work. In CVPR Workshops, pages 1783–1790, 2018. 2, 7

  19. [27]

    M. Liao, J. Zhang, Z. Wan, F. Xie, J. Liang, P. Lyu, C. Yao, and X. Bai. Scene text recognition from two-dimensional perspective. In AAAI, 2019. 8

  20. [28]

    T. Lin, P. Doll ´ar, R. B. Girshick, K. He, B. Hariharan, and S. J. Belongie. Feature pyramid networks for object detec- tion. In CVPR, pages 936–944, 2017. 2

  21. [29]

    W. Liu, C. Chen, and K. K. Wong. Char-net: A character- aware neural network for distorted scene text recognition. In AAAI, pages 7154–7161, 2018. 2, 8

  22. [30]

    Y . Liu, Z. Wang, H. Jin, and I. J. Wassell. Synthetically supervised feature learning for scene text recognition. In ECCV, pages 449–465, 2018. 2, 8

  23. [31]

    Z. Liu, Y . Li, F. Ren, W. L. Goh, and H. Yu. Squeezedtext: A real-time scene text recognition by binary convolutional encoder-decoder network. In AAAI, pages 7194–7201, 2018. 8

  24. [32]

    S. Long, X. He, and C. Yao. Scene text detection and recognition: The deep learning era. arXiv preprint arXiv:1811.04256, 2018. 1, 2

  25. [33]

    S. Long, J. Ruan, W. Zhang, X. He, W. Wu, and C. Yao. Textsnake: A flexible representation for detecting text of ar- bitrary shapes. In ECCV, pages 19–35. Springer, 2018. 3

  26. [34]

    S. M. Lucas, A. Panaretos, L. Sosa, A. Tang, S. Wong, and R. Young. Icdar 2003 robust reading competitions. In IC- DAR, page 682. IEEE, 2003. 5

  27. [35]

    P. Lyu, M. Liao, C. Yao, W. Wu, and X. Bai. Mask textspot- ter: An end-to-end trainable neural network for spotting text with arbitrary shapes. In Proceedings of the European Con- ference on Computer Vision (ECCV), pages 67–83, 2018. 1

  28. [36]

    P. Lyu, C. Yao, W. Wu, S. Yan, and X. Bai. Multi-oriented scene text detection via corner localization and region seg- mentation. In Proceedings of the IEEE Conference on Com- puter Vision and Pattern Recognition , pages 7553–7563,

  29. [37]

    Mishra, K

    A. Mishra, K. Alahari, and C. Jawahar. Scene text recogni- tion using higher order language priors. In BMVC. BMV A,

  30. [38]

    Mishra, K

    A. Mishra, K. Alahari, and C. Jawahar. Top-down and bottom-up cues for scene text recognition. In CVPR. IEEE,

  31. [39]

    Novikova, O

    T. Novikova, O. Barinova, P. Kohli, and V . S. Lempitsky. Large-lexicon attribute-consistent text recognition in natural images. In ECCV, pages 752–765, 2012. 2

  32. [40]

    Quy Phan, P

    T. Quy Phan, P. Shivakumara, S. Tian, and C. Lim Tan. Rec- ognizing text with perspective distortion in natural scenes. In ICCV, pages 569–576, 2013. 2, 5

  33. [41]

    Risnumawan, P

    A. Risnumawan, P. Shivakumara, C. S. Chan, and C. L. Tan. A robust arbitrary text detection system for natural scene im- ages. Expert Syst. Appl., 41(18):8027–8048, 2014. 5

  34. [42]

    J. A. Rodr ´ıguez-Serrano, A. Gordo, and F. Perronnin. Label embedding: A frugal baseline for text recognition. IJCV, 113(3):193–207, 2015. 8

  35. [43]

    X. Rong, C. Yi, and Y . Tian. Recognizing text-based traffic guide panels with cascaded localization network. In ECCV Workshops, pages 109–121, 2016. 1

  36. [44]

    B. Shi, X. Bai, and C. Yao. An end-to-end trainable neural network for image-based sequence recognition and its appli- cation to scene text recognition. TPAMI, 39(11):2298–2304,

  37. [45]

    B. Shi, X. Wang, P. Lyu, C. Yao, and X. Bai. Robust scene text recognition with automatic rectification. InCVPR, pages 4168–4176, 2016. 1, 2, 8

  38. [46]

    B. Shi, M. Yang, X. Wang, P. Lyu, C. Yao, and X. Bai. Aster: an attentional scene text recognizer with flexible rectifica- tion. TPAMI, 2018. 1, 2, 8

  39. [47]

    Su and S

    B. Su and S. Lu. Accurate scene text recognition based on recurrent neural network. In ACCV, pages 35–48, 2014. 8

  40. [48]

    Su and S

    B. Su and S. Lu. Accurate recognition of words in scenes without character segmentation using recurrent neural net- work. Pattern Recognition, 63:397–405, 2017. 2

  41. [49]

    K. Wang, B. Babenko, and S. J. Belongie. End-to-end scene text recognition. In ICCV, pages 1457–1464, 2011. 2, 5, 8

  42. [50]

    Wang and S

    K. Wang and S. Belongie. Word spotting in the wild. In ECCV, pages 591–604, 2010. 2

  43. [51]

    T. Wang, D. J. Wu, A. Coates, and A. Y . Ng. End-to-end text recognition with convolutional neural networks. In ICPR, pages 3304–3308, 2012. 2, 8

  44. [52]

    X. Yang, D. He, Z. Zhou, D. Kifer, and C. L. Giles. Learning to read irregular text with attention mechanisms. In IJCAI, pages 3280–3286, 2017. 2, 8

  45. [53]

    C. Yao, X. Bai, and W. Liu. A unified framework for multi- oriented text detection and recognition. IEEE Transactions on Image Processing, 23(11):4737–4749, 2014. 1

  46. [54]

    C. Yao, X. Bai, W. Liu, Y . Ma, and Z. Tu. Detecting texts of arbitrary orientations in natural images. In 2012 IEEE Con- ference on Computer Vision and Pattern Recognition, pages 1083–1090. IEEE, 2012. 1

  47. [55]

    C. Yao, X. Bai, B. Shi, and W. Liu. Strokelets: A learned multi-scale representation for scene text recognition. In CVPR, pages 4042–4049, 2014. 2, 8

  48. [56]

    M. D. Zeiler. ADADELTA: an adaptive learning rate method. CoRR, abs/1212.5701, 2012. 6

  49. [57]

    F. Zhan, S. Lu, and C. Xue. Verisimilar image synthesis for accurate detection and recognition of texts in scenes. In ECCV, pages 257–273. Springer, 2018. 2

  50. [58]

    Zhang, W

    Z. Zhang, W. Shen, C. Yao, and X. Bai. Symmetry-based text line detection in natural scenes. In CVPR, pages 2558–2567,

  51. [59]

    X. Zhou, C. Yao, H. Wen, Y . Wang, S. Zhou, W. He, and J. Liang. East: an efficient and accurate scene text detector. In Proceedings of the IEEE conference on Computer Vision and Pattern Recognition, pages 5551–5560, 2017. 1

  52. [60]

    Y . Zhu, C. Yao, and X. Bai. Scene text detection and recog- nition: Recent advances and future trends. Frontiers of Com- puter Science, 10(1):19–36, 2016. 1

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.