Deep Watershed Detector for Music Object Recognition

@inproceedings{Tuggener2018DeepWD,
  title={Deep Watershed Detector for Music Object Recognition},
  author={Lukas Tuggener and Ismail Elezi and J{\"u}rgen Schmidhuber and Thilo Stadelmann},
  booktitle={ISMIR},
  year={2018}
}
Optical Music Recognition (OMR) is an important and challenging area within music information retrieval, the accurate detection of music symbols in digital images is a core functionality of any OMR pipeline. In this paper, we introduce a novel object detection method, based on synthetic energy maps and the watershed transform, called Deep Watershed Detector (DWD). Our method is specifically tailored to deal with high resolution images that contain a large number of very small objects and is… 

Figures and Tables from this paper

State-of-the-Art Model for Music Object Recognition with Deep Learning
TLDR
This paper proposes an end-to-end detection model based on a deep convolutional neural network and feature fusion that is able to directly process the entire image and then output the symbol categories and the pitch and duration of notes.
Printed score detection based on deep learning
TLDR
The end-to-end printed music score recognition model based on YOLOv3 is designed, which adopted multi-task learning, and learns the pitch and duration classification task and note coordinates task, which improved the generalization ability of the model.
A Baseline for General Music Object Detection with Deep Learning
TLDR
A baseline for general detection of musical symbols with deep learning is presented and the first time that competing music object detection systems from the machine learning paradigm are directly compared to each other.
The DeepScoresV2 Dataset and Benchmark for Music Object Detection
TLDR
This paper improves upon the original DeepScores dataset by providing much more detailed annotations, namely annotations for 135 classes including fundamental symbols of non-fixed size and shape, increasing the number of annotated symbols by 23%; and releases two state-of-the-art baselines for Deep scoresV2 based on Faster R-CNN and the Deep Watershed Detector.
Handwritten Historical Music Recognition by Sequence-to-Sequence with Attention Mechanism
TLDR
This paper explores the adaptation of sequence-to-sequence models with attention mechanism (used in translation and handwritten text recognition) and the generation of specific synthetic data for recognizing old music scores and demonstrates that the approach is promising, especially when compared with long short-term memory neural networks.
Learning Notation Graph Construction for Full-Pipeline Optical Music Recognition
TLDR
This work forms the notation assembly stage from a set of detected primitives as a machine learning problem, modeled as a graph that stores syntactic relationships among primitives, which allows us to capture the configuration of symbols in a music-notation document.
Optical Music Recognition: State of the Art and Major Challenges
TLDR
This paper reviews relevant works in OMR, including fundamental methods and significant outcomes, and highlights different stages of the OMR pipeline, which often lack standard input and output representation and standardised evaluation.
DoReMi: First glance at a universal OMR dataset
TLDR
This paper provides a first look at DoReMi, an OMR dataset that addresses these challenges, and a baseline object detection model to assess its utility, and opens a new course of discussions that would ultimately aid that goal.
Improved Feature Pyramid Convolutional Neural Network for Effective Recognition of Music Scores
  • Lei Li
  • Computer Science
    Computational intelligence and neuroscience
  • 2022
TLDR
An improved convolutional neural network for musical score recognition has a higher recognition accuracy and a stronger generalization performance, which allows the musicalscore recognition method to be applied to more types of music score recognition scenarios, and such a recognition model has more practical value.
Improved Feature Pyramid Convolutional Neural Network for Effective Recognition of Music Scores
TLDR
The experimental results show that the proposed improved convolutional neural network for musical score recognition model has a higher recognition accuracy and a stronger generalization performance, which allows the music score recognition method to be applied to more types of musical score Recognition scenarios, and such a recognition model have more practical value.
...
...

References

SHOWING 1-10 OF 40 REFERENCES
Handwritten Music Object Detection: Open Issues and Baseline Results
TLDR
This work proposes an end-to-end trainable object detector for music symbols that is capable of detecting almost the full vocabulary of modern music notation in handwritten music scores and shows that a machine learning approach can be used to accurately detect music objects with a mean average precision of over 80%.
Towards Self-Learning Optical Music Recognition
TLDR
A new approach to solve OMR is proposed by investigating how humans read music scores and by imitating that behavior with machine learning, which can be performed at high rates of confidence which is comparable to the performance of humans on the same task.
DeepScores-A Dataset for Segmentation, Detection and Classification of Tiny Objects
TLDR
A detailed statistical analysis of the DeepScores dataset is presented, comparing it with other computer vision datasets like PASCAL VOC, SUN, SVHN, ImageNet, MS-COCO, as well as with other OMR datasets.
Detecting Noteheads in Handwritten Scores with ConvNets and Bounding Box Regression
TLDR
This work presents ongoing work on a simple notehead detector using convolutional neural networks for pixel classification and bounding box regression that achieves a detection f-score of 0.97 on binary score images in the MUSCIMA++ dataset, does not require staff removal, and is applicable to a variety of handwriting styles and levels of musical complexity.
The MUSCIMA++ Dataset for Handwritten Optical Music Recognition
  • Jan Hajic, Pavel Pecina
  • Computer Science
    2017 14th IAPR International Conference on Document Analysis and Recognition (ICDAR)
  • 2017
TLDR
Ground truth in MUSCIMA++ is a notation graph, which analysis shows to be a necessary and sufficient representation of music notation, and is designed and collected, a new dataset for OMR.
The Challenge of Optical Music Recognition
TLDR
The challenges posed by optical music recognition are described, and a generalised framework for software is presented that emphasises key stages that must be solved: staff line identification, musical object location, musical feature classification, and musical semantics.
End-to-End Optical Music Recognition Using Neural Networks
TLDR
Results obtained depict classification error rates around 2 % at symbol level, thus proving the potential of the proposed end-to-end architecture for OMR.
Towards the Recognition of Compound Music Notes in Handwritten Music Scores
TLDR
This work proposes a method based on perceptual grouping for the recognition of compound music notes and has been tested using several handwritten music scores of the CVC-MUSCIMA database and compared with a commercial Optical Music Recognition software.
Staff-line removal with selectional auto-encoders
Bootstrapping Samples of Accidentals in Dense Piano Scores for CNN-Based Detection
TLDR
A data augmentation bootstrapping method is used to accurately train a deep learning model to do the localization and classification of an accidental symbol associated with a note head, or the note head if there is no accidental.
...
...