Music staff removal with supervised pixel classification

Calvo-Zaragoza, Jorge; Micó, Luisa; Oncina, José

doi:10.1007/s10032-016-0266-2

Cited by 18 publications

(14 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…With DocCreator any DIAR researcher can create complete groundtruthed images and increase the size of its document image database. After 5 years of several collaborations and test campaigns, DocCreator (or databases created with DocCreator) have been tested by different researchers and used to publish [25], [26], [27], [28] proving its utility for performance evaluation or retraining tasks. DocCreator is on an open source ongoing project.…”

Section: Discussionmentioning

confidence: 99%

Massive, Free and Reproducible Grountruthed Document Image Databases Generation with DocCreator

Journet¹,

Mansencal

Visani

2017

2017 14th IAPR International Conference on Document Analysis and Recognition (ICDAR)

View full text Add to dashboard Cite

Whether your research is focused on image restoration, layout analysis, text-graphic separation, binarization, OCR, etc. you need a groundtruthed database to train your method or to evaluate it. This article presents DocCreator, a multi-platform and open-source software able to create many synthetic image documents with controlled groundtruth. With DocCreator, you can create complete synthetic images choosing the text, font, background and layout to use, add various realistic degradations (bleedthrough, light defect, paper deformation, ink degradation, etc.) on original images, or combine both to increase the size of your database. DocCreator comes as an online (easy to test version) and a desktop solution (fast calculation process, and no need to upload copyrighted data). DocCreator is useful for retraining tasks and to know precisely whether your algorithm is robust. It has already been used favorably and could help other DIAR researchers to produce and share groundtruthed databases.

show abstract

Section: Discussionmentioning

confidence: 99%

Massive, Free and Reproducible Grountruthed Document Image Databases Generation with DocCreator

Journet¹,

Mansencal

Visani

2017

2017 14th IAPR International Conference on Document Analysis and Recognition (ICDAR)

View full text Add to dashboard Cite

show abstract

“…The comparison with Pixel method is also illustrative of the goodness of our proposal, since it demonstrates that the performance is not only achieved by using a supervised learning scheme (Pixel also does so) but because of the adequacy of the proposed model. [21] 83.01 2013 NUASI lin [3] 94.29 2008 NUASI skel [3] 93.34 2008 LRDE [11] 97.14 2014 INESC [10] 91.01 2009 NUS [9] 75.24 2012 Pixel [22] 95.04 2016 Image Operator [12] 97.96 2017 StaffNet [26] 97.87 2017 Baseline 97.31 -Our approach 99.32 - We also test the method using grayscale version of the scores. Our approach can easily be used to deal with grayscale images without any additional pre-processing steps like binarization.…”

Section: E Comparison With State-of-the-artmentioning

confidence: 99%

Staff line Removal using Generative Adversarial Networks

Konwer

Bhunia

Bhowmick

et al. 2018

2018 24th International Conference on Pattern Recognition (ICPR)

View full text Add to dashboard Cite

Staff line removal is a crucial pre-processing step in Optical Music Recognition. It is a challenging task to simultaneously reduce the noise and also retain the quality of music symbol context in ancient degraded music score images. In this paper we propose a novel approach for staff line removal, based on Generative Adversarial Networks. We convert staff line images into patches and feed them into a U-Net, used as Generator. The Generator intends to produce staff-less images at the output. Then the Discriminator does binary classification and differentiates between the generated fake staff-less image and real ground truth staff less image. For training, we use a Loss function which is a weighted combination of L2 loss and Adversarial loss. L2 loss minimizes the difference between real and fake staff-less image. Adversarial loss helps to retrieve more high quality textures in generated images. Thus our architecture supports solutions which are closer to ground truth and it reflects in our results. For evaluation we consider the ICDAR/GREC 2013 staff removal database. Our method achieves superior performance in comparison to other conventional approaches.

show abstract

“…To address these issues, we consider the problem of removing staff-lines from music score images. This problem has been receiving attention recently [16], [17], [18] and there is a large public dataset with ground-truth information [16], making it interesting to our study. Moreover, there are performance reports of heuristic methods and also of learned image operators [9], which will allow us to make a direct comparison.…”

Section: Introductionmentioning

confidence: 99%

Image Operator Learning Coupled with CNN Classification and Its Application to Staff Line Removal

Aguilar

Hirata

2017

2017 14th IAPR International Conference on Document Analysis and Recognition (ICDAR)

View full text Add to dashboard Cite

Many image transformations can be modeled by image operators that are characterized by pixel-wise local functions defined on a finite support window. In image operator learning, these functions are estimated from training data using machine learning techniques. Input size is usually a critical issue when using learning algorithms, and it limits the size of practicable windows. We propose the use of convolutional neural networks (CNNs) to overcome this limitation. The problem of removing staff-lines in music score images is chosen to evaluate the effects of window and convolutional mask sizes on the learned image operator performance. Results show that the CNN based solution outperforms previous ones obtained using conventional learning algorithms or heuristic algorithms, indicating the potential of CNNs as base classifiers in image operator learning. The implementations will be made available on the TRIOSlib project site.

show abstract

Music staff removal with supervised pixel classification

Cited by 18 publications

References 31 publications

Massive, Free and Reproducible Grountruthed Document Image Databases Generation with DocCreator

Massive, Free and Reproducible Grountruthed Document Image Databases Generation with DocCreator

Staff line Removal using Generative Adversarial Networks

Image Operator Learning Coupled with CNN Classification and Its Application to Staff Line Removal

Contact Info

Product

Resources

About