Pith. sign in

REVIEW 4 major objections 7 minor 46 references

Synthesising Handwritten Music with GANs: A Comprehensive Evaluation of CycleWGAN, ProGAN, and DCGAN

T0 review · 4 major / 7 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper claims that CycleWGAN, a CycleGAN variant with Wasserstein loss, outperforms DCGAN and ProGAN at generating handwritten music crops, scoring FID 41.87, IS 2.29, and KID 0.05.

desk verdict Useful benchmark of GAN variants for handwritten music, but the headline 'CycleWGAN significantly outperforms' is confounded by a conditional-vs-unconditional task mismatch and a 200-image evaluation set. read the letter →

arxiv 2411.16405 v1 pith:5HLET42V submitted 2024-11-25 cs.CV cs.AIeess.IV

Elona Shatri , Kalikidhar Palavala , George Fazekas This is my paper · ORCID
classification cs.CVcs.AIeess.IV
keywords handwrittenmusicgenerationopticalrecognitionCycleWGANgenerativeadversarialnetworksimage-to-imagetranslationWassersteinGANFIDandKIDevaluationdataaugmentation
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 addresses the scarcity of annotated handwritten music for Optical Music Recognition by testing whether GANs can generate convincing handwritten score crops. It compares three architectures and argues that CycleWGAN, a CycleGAN modified with Wasserstein loss, instance normalisation, and residual blocks, produces the most realistic and diverse images of the three. On 200 evaluation images, CycleWGAN reaches FID 41.87, IS 2.29, and KID 0.05, beating ProGAN (FID 49.54) and DCGAN (FID 210.18). A sympathetic reader would take the central claim as: for crop-level handwritten music synthesis, the Wasserstein-stabilised cycle-consistency architecture is the strongest of the tested options and a better baseline than the earlier CycleGAN. The paper also concedes remaining flaws, such as gaps in note-head transfer, that keep the generated images from direct OMR training.

What carries the argument

The load-bearing object is the CycleWGAN architecture: two generators and two discriminators translating between printed and handwritten music domains, trained with Wasserstein adversarial losses plus a cycle-consistency loss that forces round trips to reconstruct the input. Supporting machinery includes instance normalisation, chosen because batch normalisation is incompatible with Wasserstein training, nine ResNet blocks in the generator, DCGAN as a 64x64 baseline, and ProGAN with progressive growth from 4x4 to 128x128 and WGAN-GP. The comparison is carried by standard generative metrics, FID, IS, and KID, supplemented by qualitative object detection and PCA.

What would settle it

Compute FID, IS, and KID on at least a few thousand generated images with bootstrap confidence intervals; the CycleWGAN claim collapses if its FID advantage over ProGAN falls inside the noise band. A second falsifier is a downstream OMR test: if a recogniser trained with CycleWGAN crops does not match or beat training on ProGAN crops or real data, the practical claim for improving OMR is unsupported.

Watch

Extended reading notes

Core claim

CycleWGAN replaces the standard adversarial losses in CycleGAN with the Wasserstein distance, uses instance normalisation in both networks, and employs a ResNet-based generator with nine residual blocks while keeping an L2 cycle-consistency loss with lambda equal to 10. The paper's finding is that these changes improve training stability and style transfer enough that CycleWGAN substantially outperforms DCGAN and ProGAN on FID, IS, and KID, and that PCA places its outputs closest to real handwritten crops. The paper presents this as evidence that CycleWGAN is a promising data-augmentation tool for OMR, while noting that symbol-level errors, such as imperfect note heads and accidentals, mean the images are not yet ready for direct use in training OMR systems.

Load-bearing premise

The ranking rests on the assumption that FID, IS, and KID computed on 200 generated crops are stable enough to separate the models, despite the paper's own note that FID typically needs larger samples and the absence of confidence intervals or significance tests.

Editorial extensions

If this is right

  • CycleWGAN provides a stronger baseline than the earlier CycleGAN for handwriting style transfer in music, so future image-to-image work can start from the Wasserstein variant.
  • The generated crops, after correcting note-head and accidental errors, could augment OMR training sets and reduce the annotation bottleneck.
  • ProGAN's greater symbol completeness and CycleWGAN's superior style closeness are complementary, suggesting that combining or conditioning these approaches is a plausible next step.
  • The reported metric margins need validation on larger evaluation sets before being used as a design decision for OMR pipelines, as the paper itself notes FID typically requires larger samples.
  • If CycleWGAN is right, data-hungry OMR systems can benefit from synthetic handwritten crops rather than relying only on scarce archives.

Reading between the lines

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

  • Editorial inference: the 200-image evaluation set makes FID the least reliable of the three metrics; KID's unbiased estimator is more believable at this sample size, and its large gap (0.05 vs 0.17) is the strongest quantitative evidence for CycleWGAN.
  • Editorial inference: because the models are trained and evaluated on square crops, the ranking may not survive the harder task of generating full pages; a full-page or stream-based evaluation could invert the ordering.
  • Editorial inference: a direct downstream test, training an OMR recogniser on real plus synthetic crops and measuring symbol error rate on held-out manuscripts, would settle practical value faster than FID, IS, or KID.
  • Editorial inference: incorporating WGAN-GP into CycleWGAN, as the paper suggests, is a concrete next experiment; the ProGAN results show gradient penalty helps stability, but it may change CycleWGAN's style-transfer behaviour.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 7 minor

Summary. The paper applies three GAN architectures—DCGAN, ProGAN, and CycleWGAN (a CycleGAN variant using Wasserstein losses and instance normalization)—to synthesize handwritten music crops from the CVC-MUSCIMA and DoReMi datasets. It reports quantitative metrics (FID, IS, KID) and qualitative analyses including object detection and PCA, and claims that CycleWGAN significantly outperforms the other two models and offers a promising route to improving OMR training data. The paper also provides hyperparameter details and a discussion of limitations.

Significance. If the comparison were task-equivalent and statistically grounded, this would be a useful empirical baseline for GAN-based handwritten music synthesis. The CycleWGAN adaptation is a reasonable engineering contribution, and the qualitative evaluation with a pre-trained object detector is a valuable addition. However, the central claim is not supported as stated: the comparison confounds conditioning with generation ability, the quantitative metrics lack uncertainty quantification, and the paper's own qualitative results contradict the headline conclusion. The work is nevertheless a reasonable starting point for a more careful evaluation.

major comments (4)
  1. [Sec. III-B1/2/3; Sec. V-B, Table I] The comparison is not task-equivalent. CycleWGAN is a conditional image-to-image translator that takes printed music crops as input, while DCGAN and ProGAN are unconditional generators that map latent noise z to images. The lower FID of CycleWGAN (41.87 vs 49.54 vs 210.18) may therefore reflect the strong conditioning signal rather than superior generative ability. To support the abstract's claim of 'significantly outperforms', the evaluation must either condition all models on the same printed input or evaluate CycleWGAN in an unconditional setting; at a minimum, the claims should be re-framed as 'best on the conditional translation task'.
  2. [Sec. V-B, Table I; Sec. V-D] The quantitative ranking relies on point estimates from a 200-image evaluation set with no confidence intervals or significance tests. The Inception Scores overlap within one standard deviation (CycleWGAN 2.29 ± 0.19, DCGAN 2.17 ± 0.14, ProGAN 2.01 ± 0.20), so the word 'significantly' is not supported. The paper also does not specify how the 200 evaluation images were chosen, whether the same real-image reference set was used for all models, or how many generated images were evaluated. Please provide confidence intervals (e.g., bootstrap) or a significance test, and describe the evaluation set construction; otherwise the reported margins may be evaluation noise.
  3. [Sec. V-A; Sec. VI] The qualitative results contradict the central claim. The paper states that CycleWGAN outputs contain gaps, pitch shifts, and missing noteheads and are 'unsuitable for direct use in training OMR systems', while ProGAN produces 'more complete and balanced images'. This internal contradiction needs to be resolved: either the quantitative advantage of CycleWGAN is not diagnostically meaningful for the stated application, or the conclusion that CycleWGAN is 'a promising solution for improving OMR systems' should be substantially weakened.
  4. [Sec. V-B, Eq. (5)] The Inception Score is computed with an ImageNet-pretrained Inception network, but handwritten music crops are far outside that domain, making the conditional label distribution p(y|x) largely arbitrary and the IS values difficult to interpret. The paper itself concedes that the direct applicability of these metrics 'is less established', yet IS is used as headline evidence. Either justify the use of IS for this domain (e.g., with a domain-adapted classifier) or remove it from the main claims and rely on FID/KID plus the qualitative evaluation.
minor comments (7)
  1. [Sec. V-B] The text contains the literal placeholder '[cite key papers]' in the paragraph motivating metric selection; this must be completed before submission.
  2. [Sec. V-A] The statement that DCGAN suffers from mode collapse '(see Figure 6(c))' is incorrect because Figure 6(c) is the printed input image, not a DCGAN output; please fix the cross-reference and check all figure references.
  3. [Sec. V-C, Fig. 9] The text describes 'Figure 9(b)' as comparing real handwritten, real printed, and CycleWGAN-generated images, but in the figure caption panel (b) is the ProGAN panel; correct either the text or the caption.
  4. [Throughout] Capitalization of model names is inconsistent in the text, Table I, and figures (DCGAN, PROGAN, ProGAN, CycleWGAN); please standardize.
  5. [References] Reference [26], cited as 'Improved CycleGAN for image-to-image translation, 2021', lacks author names and venue; provide full bibliographic details.
  6. [Sec. IV-C] The phrase 'Experimented with custom beta values' should read 'We experimented with custom beta values'.
  7. [Sec. III-A] The text refers to 'Figures 1a and 1b' but the sub-caption labels in Figure 1 are not defined in the caption; adding explicit '(a)' and '(b)' labels would improve clarity.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the paper is an empirical GAN comparison with measured metrics, and its self-citations are not load-bearing.

full rationale

The paper's central claim is an empirical ranking of CycleWGAN, ProGAN, and DCGAN based on measured IS, FID, and KID values reported in Section V-B and Table I. These numbers are outputs of an evaluation procedure, not quantities forced by the model definitions or loss equations in Section III-B. The CycleWGAN loss functions (Wasserstein adversarial losses plus cycle consistency) do not algebraically determine the FID/IS/KID values, and no fitted parameter is renamed as a prediction. The self-citations, most notably Ref. [7] for an earlier CycleGAN baseline, are used for comparison and for motivating inpainting-style refinements, but the CycleWGAN result does not reduce to a claim imported from that reference. The paper even flags a relevant external-validity limitation in Section V-D: the evaluation uses only 200 images, and FID typically requires larger sample sizes; this weakens the strength of the ranking but is not a circularity. The possible task mismatch between the conditional image-to-image model (CycleWGAN) and unconditional generators (DCGAN, ProGAN) is a substantive fairness-of-comparison concern, but it does not make the reported measurements equivalent to the model's inputs by construction. Overall, the derivation chain here is empirical and self-contained, so there is no significant circularity.

Assumptions & free parameters 6 free parameters · 4 assumptions · 1 invented entities

The claims rest on standard GAN training assumptions plus several hand-set hyperparameters, a small 200-image evaluation set, and the assumption that Inception-based metrics transfer to music notation. None of the hyperparameters are fitted to the target metrics in a formal sense, but the reported ranking is conditional on these choices. CycleWGAN itself is a composition of known components rather than a newly invented physical entity.

free parameters (6)
  • Cycle consistency weight lambda_cycle = 10
    Chosen from CycleGAN literature (Ref. [26]) and fixed during training; the central CycleWGAN comparison depends on it.
  • Gradient penalty coefficient lambda_GP (ProGAN) = 10
    Taken from WGAN-GP (Ref. [22]); affects ProGAN training stability and thus its FID/IS/KID.
  • Learning rates = DCGAN 2e-4, ProGAN 1e-3, CycleWGAN 1e-5
    Hand-set per model; the reported performance ranking depends on these choices, and no sweep is reported.
  • ProGAN latent dimension z = 256
    Reduced from 512 to 256 to fit memory; affects output diversity.
  • CycleWGAN residual blocks = 9
    Architecture choice said to capture handwriting variation; no ablation is reported.
  • Evaluation set size = 200 images
    All FID/IS/KID values are computed on 200 samples; the paper notes this is small and may make FID unreliable.
assumptions (4)
  • domain assumption CVC-MUSCIMA and DoReMi crops form a valid paired/unpaired training set for printed-to-handwritten music translation.
    Sec III-A; if these datasets are not representative, the models learn a narrow style, but this is a standard domain assumption.
  • domain assumption Inception-network features used by FID/IS/KID are meaningful for handwritten music images.
    Sec V-B; the paper notes the applicability of these metrics to music is 'less established', yet uses them as primary quantitative evidence.
  • domain assumption The 200-image evaluation set gives stable estimates of FID/IS/KID.
    Sec V-D admits FID typically needs larger samples; the ranking could change with more samples.
  • domain assumption Cycle-consistency loss with L2 norm preserves musical content during translation.
    Sec III-C; no independent verification that pitch and rhythm are preserved, and qualitative results show notehead gaps and pitch shifts.
invented entities (1)
  • CycleWGAN architecture variant
    purpose: Hybrid of CycleGAN and Wasserstein GAN for printed-to-handwritten music translation
    The paper presents CycleWGAN as its proposed model, but provides no theoretical guarantee or external validation beyond metrics on one dataset; it is an assembly of known components (CycleGAN, WGAN loss, instance norm, residual blocks).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Synthesising Handwritten Music with GANs: A Comprehensive Evaluation of CycleWGAN, ProGAN, and DCGAN." pith.science (2026). https://pith.science/paper/5HLET42V

@misc{pith2026241116405,
  author       = {Pith},
  title        = {Pith review of: Synthesising Handwritten Music with GANs: A Comprehensive Evaluation of CycleWGAN, ProGAN, and DCGAN},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5HLET42V}},
  note         = {Machine review of arXiv:2411.16405}
}
read the original abstract

The generation of handwritten music sheets is a crucial step toward enhancing Optical Music Recognition (OMR) systems, which rely on large and diverse datasets for optimal performance. However, handwritten music sheets, often found in archives, present challenges for digitisation due to their fragility, varied handwriting styles, and image quality. This paper addresses the data scarcity problem by applying Generative Adversarial Networks (GANs) to synthesise realistic handwritten music sheets. We provide a comprehensive evaluation of three GAN models - DCGAN, ProGAN, and CycleWGAN - comparing their ability to generate diverse and high-quality handwritten music images. The proposed CycleWGAN model, which enhances style transfer and training stability, significantly outperforms DCGAN and ProGAN in both qualitative and quantitative evaluations. CycleWGAN achieves superior performance, with an FID score of 41.87, an IS of 2.29, and a KID of 0.05, making it a promising solution for improving OMR systems.

Figures

Figures reproduced from arXiv: 2411.16405 by the authors.

Figure 1
Figure 1. (a) Handwritten image crop from the CVC-MUSCIMA [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 3
Figure 3. Training loss of DCGAN over 100 epochs. 3) CycleWGAN: Improved Style Transfer with Wasserstein Loss: To further improve the generation of handwritten music images, we employ CycleWGAN, a modified version of Cycle￾GAN [26]. While CycleGAN is effective for unpaired image￾to-image translation, it can suffer from unstable training and mode collapse when dealing with complex image structures, such as musical symbols. Cyc… view at source ↗
Figure 2
Figure 2. CycleWGAN pipeline architecture used for handwritten and printed scores. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figures from the paper (5 more)
Figure 5
Figure 5. Figure 5: Training loss of (Gen H, Gen P) and (Disc H, Disc P) [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 4
Figure 4. Figure 4: Training loss of ProGAN over 180 epochs demonstrat [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 6
Figure 6. Figure 6: (a) CycleWGAN-generated handwritten image. (b) [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 8
Figure 8. Figure 8: Printed music images and their corresponding images [PITH_FULL_IMAGE:figures/full_fig_p007_8.png]
Figure 9
Figure 9. Figure 9: PCA analysis of test data for individual GAN models [PITH_FULL_IMAGE:figures/full_fig_p009_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

46 extracted references · 36 canonical work pages

  1. [1]

    Haji ˇc and P

    J. Haji ˇc and P. Pecina, ”The MUSCIMA++ Dataset for Handwritten Optical Music Recognition,” in Proc. 14th IAPR Int. Conf. Document Analysis and Recognition (ICDAR) , Kyoto, Japan, 2017, pp. 39–46, doi: 10.1109/ICDAR.2017.16

    doi:10.1109/icdar.2017.16 2017

  2. [2]

    Bar ´o, P

    A. Bar ´o, P. Riba, J. Calvo-Zaragoza, and A. Forn ´es, ”From Optical Music Recognition to Handwritten Music Recognition: A baseline,” Pattern Recognition Letters , vol. 123, 2019

    work page 2019

  3. [3]

    J. Y . Zhu, T. Park, P. Isola, and A. A. Efros, ”Unpaired image-to-image translation using cycle-consistent adversarial networks,” in Proc. IEEE Int. Conf. Computer Vision (ICCV) , 2017, pp. 2223–2232

    work page 2017

  4. [4]

    Karras, S

    T. Karras, S. Laine, and T. Aila, ”A style-based generator architecture for generative adversarial networks,” in Proc. IEEE/CVF Conf. Computer Vision and Pattern Recognition (CVPR) , 2019, pp. 4401–4410

    work page 2019

  5. [5]

    Shatri and G

    E. Shatri and G. Fazekas, ”Optical music recognition: State of the art and major challenges,” arXiv preprint, arXiv:2006.07885, 2020

    arXiv 2006

  6. [6]

    Shatri and G

    E. Shatri and G. Fazekas, ”DoReMi: First glance at a universal OMR dataset,” arXiv preprint, arXiv:2107.07786, Jul. 2021

    arXiv 2021

  7. [7]

    Hande, E

    P. Hande, E. Shatri, B. Timms, and G. Fazekas, ”Towards Artificially Generated Handwritten Sheet Music Datasets,” in Proc. 5th Int. Work- shop on Reading Music Systems , 2023, p. 25

    work page 2023

  8. [8]

    Calvo-Zaragoza, J

    J. Calvo-Zaragoza, J. H. Jr., and A. Pacha, ”Understanding optical music recognition,” ACM Comput. Surveys (CSUR) , vol. 53, no. 4, pp. 1–35, 2020

    work page 2020

Show all 46 references
  1. [9]

    Pacha, K

    A. Pacha, K. Y . Choi, B. Cou ¨asnon, Y . Ricquebourg, R. Zanibbi, and H. Eidenberger, ”Handwritten music object detection: Open issues and baseline results,” in Proc. 13th IAPR Int. Workshop on Document Analysis Systems (DAS) , 2018, pp. 163–168

    2018

  2. [10]

    S. Ren, K. He, R. Girshick, and J. Sun, ”Faster R-CNN: Towards real-time object detection with region proposal networks,” Advances in Neural Information Processing Systems (NeurIPS) , vol. 28, 2015

    2015

  3. [11]

    Z. Wang, Q. She, and T. E. Ward, ”Generative Adversarial Networks in Computer Vision: A Survey and Taxonomy,” ACM Comput. Surveys , vol. 54, no. 2, Article 37, pp. 1–38, Mar. 2022, doi: 10.1145/3439723

    2022 doi

  4. [12]

    J. Dai, Y . Li, K. He, and J. Sun, ”R-FCN: Object detection via region- based fully convolutional networks,” in Advances in Neural Information Processing Systems (NeurIPS) , vol. 29, 2016

    2016

  5. [13]

    W. Liu, D. Anguelov, D. Erhan, C. Szegedy, S. Reed, C. Y . Fu, and A. C. Berg, ”SSD: Single shot multibox detector,” in Proc. Eur . Conf. Computer Vision (ECCV) , Amsterdam, The Netherlands, 2016, pp. 21– 37, Springer International Publishing

    2016

  6. [14]

    Mirza and S

    M. Mirza and S. Osindero, ”Conditional generative adversarial nets,” arXiv preprint, arXiv:1411.1784, 2014

    2014 arXiv

  7. [15]

    Cardoso, S

    R. Cardoso, S. Vallecorsa, and E. Nemni, ”Conditional progressive generative adversarial network for satellite image generation,” arXiv preprint, arXiv:2211.15303, Nov. 2022

    2022 arXiv

  8. [16]

    L. Sun, J. Chen, Y . Xu, M. Gong, K. Yu, and K. Batmanghelich, ”Hierarchical amortized GAN for 3D high-resolution medical image synthesis,” IEEE J. Biomed. Health Informatics , vol. 26, no. 8, pp. 3966– 3975, May 2022

    2022

  9. [17]

    F. F. Farooqui, M. Hassan, M. S. Younis, and M. K. Siddhu, ”Offline hand-written Urdu word spotting using random data generation,” IEEE Access, vol. 8, pp. 131119–131136, Jul. 2020

    2020

  10. [18]

    R ´ıos-Vila, J

    A. R ´ıos-Vila, J. M. I ˜nesta, and J. Calvo-Zaragoza, ”On the use of transformers for end-to-end optical music recognition,” in Proc. Iberian Conf. Pattern Recognition and Image Analysis , Apr. 2022, pp. 470–481, Cham: Springer International Publishing

    2022

  11. [19]

    Huang, X

    Z. Huang, X. Jia, and Y . Guo, ”State-of-the-art model for music object recognition with deep learning,” Applied Sciences , vol. 9, no. 13, p. 2645, Jun. 2019

    2019

  12. [20]

    Bar ´o, P

    A. Bar ´o, P. Riba, and A. Forn´es, ”Towards the recognition of compound music notes in handwritten music scores,” in Proc. 15th Int. Conf. Frontiers in Handwriting Recognition (ICFHR) , Oct. 2016, pp. 465– 470

    2016

  13. [21]

    F. A. Croitoru, V . Hondru, R. T. Ionescu, and M. Shah, ”Diffusion models in vision: A survey,” IEEE Trans. Pattern Anal. Mach. Intell. , vol. 45, no. 9, pp. 10850–10869, Mar. 2023

    2023

  14. [22]

    Gulrajani, F

    I. Gulrajani, F. Ahmed, M. Arjovsky, V . Dumoulin, and A. C. Courville, ”Improved training of Wasserstein GANs,” in Advances in Neural Information Processing Systems (NeurIPS) , vol. 30, 2017

    2017

  15. [23]

    Ulyanov, A

    D. Ulyanov, A. Vedaldi, and V . Lempitsky, ”Instance normaliza- tion: The missing ingredient for fast stylization,” arXiv preprint , arXiv:1607.08022, Jul. 2016

    2016 arXiv

  16. [24]

    Arjovsky, S

    M. Arjovsky, S. Chintala, and L. Bottou, ”Wasserstein generative ad- versarial networks,” in Proc. Int. Conf. Machine Learning (ICML) , Jul. 2017, pp. 214–223

    2017

  17. [25]

    Gemp and B

    I. Gemp and B. McWilliams, ”The unreasonable effectiveness of Adam on cycles,” in Proc. NeurIPS Workshop Bridging Game Theory and Deep Learning, 2019

    2019

  18. [26]

    W. Hu, M. Li, and X. Ju, ”Improved CycleGAN for image-to-image translation,” 2021

    2021

  19. [27]

    Parmar, R

    G. Parmar, R. Zhang, and J. Y . Zhu, ”On aliased resizing and surprising subtleties in GAN evaluation,” in Proc. IEEE/CVF Conf. Computer Vision and Pattern Recognition (CVPR) , 2022, pp. 11410–11420

    2022

  20. [28]

    Salimans, I

    T. Salimans, I. Goodfellow, W. Zaremba, V . Cheung, A. Radford, and X. Chen, ”Improved techniques for training GANs,” in Advances in Neural Information Processing Systems (NeurIPS) , vol. 29, 2016

    2016

  21. [29]

    Szegedy, V

    C. Szegedy, V . Vanhoucke, S. Ioffe, J. Shlens, and Z. Wojna, ”Rethinking the inception architecture for computer vision,” in Proc. IEEE Conf. Computer Vision and Pattern Recognition (CVPR) , 2016, pp. 2818– 2826

    2016

  22. [30]

    Borji, ”Pros and cons of GAN evaluation measures: New develop- ments,” Computer Vision and Image Understanding , vol

    A. Borji, ”Pros and cons of GAN evaluation measures: New develop- ments,” Computer Vision and Image Understanding , vol. 215, Jan. 2022, Art. no. 103329

    2022

  23. [31]

    Heusel, H

    M. Heusel, H. Ramsauer, T. Unterthiner, B. Nessler, and S. Hochreiter, ”GANs trained by a two-time-scale update rule converge to a local Nash equilibrium,” in Advances in Neural Information Processing Systems (NeurIPS), vol. 30, 2017

    2017

  24. [32]

    K. He, X. Zhang, S. Ren, and J. Sun, ”Deep residual learning for image recognition,” in Proc. IEEE Conf. Computer Vision and Pattern Recognition (CVPR) , 2016, pp. 770–778

    2016

  25. [33]

    Bi ´nkowski, D

    M. Bi ´nkowski, D. J. Sutherland, M. Arbel, and A. Gretton, ”Demysti- fying MMD GANs,” arXiv preprint, arXiv:1801.01401, Jan. 2018

    2018 arXiv

  26. [34]

    Havelka, J

    J. Havelka, J. Mayer, and P. Pecina, ”Symbol Generation via Autoen- coders for Handwritten Music Synthesis,” in Proc. 5th Int. Workshop on Reading Music Systems , Nov. 2023, p. 20

    2023

  27. [35]

    Mayer and P

    J. Mayer and P. Pecina, ”Synthesizing training data for handwritten music recognition,” in Proc. 16th Int. Conf. Document Analysis and Recognition (ICDAR), Lausanne, Switzerland, Sep. 2021, pp. 626–641, Springer International Publishing

    2021

  28. [36]

    Radford, ”Unsupervised representation learning with deep convolutional generative adversarial networks,” arXiv preprint , arXiv:1511.06434, 2015

    A. Radford, ”Unsupervised representation learning with deep convolutional generative adversarial networks,” arXiv preprint , arXiv:1511.06434, 2015

    2015 arXiv

  29. [37]

    Goodfellow, J

    I. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, and Y . Bengio, ”Generative adversarial nets,” in Advances in Neural Information Processing Systems (NeurIPS) , vol. 27, 2014

    2014

  30. [38]

    R ´ıos-Vila, J

    A. R ´ıos-Vila, J. Calvo-Zaragoza, and T. Paquet, ”Sheet Music Trans- former: End-To-End Optical Music Recognition Beyond Monophonic Transcription,” arXiv preprint, arXiv:2402.07596, Feb. 2024

    2024 arXiv

  31. [39]

    van Der Wel and K

    E. van Der Wel and K. Ullrich, ”Optical music recognition with convolu- tional sequence-to-sequence models,” arXiv preprint, arXiv:1707.04877, Jul. 2017

    2017 arXiv

  32. [40]

    Forn ´es, A

    A. Forn ´es, A. Dutta, A. Gordo, and J. Llad ´os, ”CVC-MUSCIMA: A ground truth of handwritten music score images for writer identification and staff removal,” Int. J. Document Anal. Recognit. (IJDAR) , vol. 15, pp. 243–251, Sep. 2012

    2012

  33. [41]

    Tirupati, E

    N. Tirupati, E. Shatri, and G. Fazekas, ”Crafting Handwritten Notations: Towards Sheet Music Generation,” in To appear in Proc. 6th Int. Workshop on Reading Music Systems , 2024

    2024

  34. [42]

    Bar ´o, C

    A. Bar ´o, C. Badal, and A. Forn ´es, ”Handwritten historical music recog- nition by sequence-to-sequence with attention mechanism,” in Proc. 17th Int. Conf. Frontiers in Handwriting Recognition (ICFHR), Sep. 2020, pp. 205–210

    2020

  35. [43]

    I. B. Mustapha, S. Hasan, H. Nabus, and S. M. Shamsuddin, ”Condi- tional deep convolutional generative adversarial networks for isolated handwritten Arabic character generation,” Arabian J. Sci. Eng. , vol. 47, no. 2, pp. 1309–1320, Feb. 2022

    2022

  36. [44]

    Calvo-Zaragoza and D

    J. Calvo-Zaragoza and D. Rizo, ”End-to-end neural optical music recognition of monophonic scores,” Applied Sciences , vol. 8, no. 4, p. 606, Apr. 2018

    2018

  37. [45]

    Gretton, K

    A. Gretton, K. M. Borgwardt, M. J. Rasch, B. Sch ¨olkopf, and A. Smola, ”A kernel two-sample test,” J. Mach. Learn. Res. , vol. 13, no. 1, pp. 723–773, Mar. 2012

    2012

  38. [46]

    Shatri and G

    E. Shatri and G. Fazekas, ”Knowledge Discovery in Optical Music Recognition: Enhancing Information Retrieval with Instance Segmenta- tion,” in Proc. Int. Conf. Knowledge Discovery and Information Retrieval (KDIR), 2024

    2024

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.