CNN-Based Accidental Detection in Dense Printed Piano Scores - Archive ouverte HAL Access content directly
Conference Papers Year :

CNN-Based Accidental Detection in Dense Printed Piano Scores

Kwon-Young Choi
Richard Zanibbi
  • Function : Author
  • PersonId : 907998


The recognition of mid-18th to mid-20th century piano scores presents segmentation challenges caused by touching and broken symbols produced by imprinting techniques and time degradation. We present a new notehead accidental dataset containing 2955 images from dense and damaged piano scores. We address this detection problem with very small training samples using a simple Spatial Transformer (ST)-based Convolutional Neural Network detector improved through bootstrapping and contextual information, and more powerful deep learning detectors (Faster R-CNN, R-FCN, and SSD) with transfer-learning on the COCO dataset. We trained all our detectors using 5 fold cross-validation and obtain 98.73% mean Average Precision (mAP) for an Intersection over Union (IoU) threshold of 0.75 with our best detector. Our ST-based detector obtains a slightly lower mAP of 94.81%, but runs 40 times faster, and uses 18 times less memory.
Fichier principal
Vignette du fichier
camera_ready_icdar_2019.pdf (413.29 Ko) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-02430041 , version 1 (07-01-2020)


  • HAL Id : hal-02430041 , version 1


Kwon-Young Choi, Bertrand B. Couasnon, Yann Ricquebourg, Richard Zanibbi. CNN-Based Accidental Detection in Dense Printed Piano Scores. 15th International Conference on Document Analysis and Recognition, Sep 2019, Sydney, Australia. ⟨hal-02430041⟩
60 View
200 Download


Gmail Facebook Twitter LinkedIn More