MUSCIMA++
收藏资源简介:
MUSCIMA++是由查尔斯大学数学与物理学院形式与应用语言学研究所创建的一个手写音乐符号数据集,包含140页手写音乐,共计91255个手动标注的音乐符号和82261个符号对之间的明确关系。该数据集支持训练和评估符号分类、符号定位和音乐图谱组装模型,适用于单独和联合任务。数据集提供了开源工具,用于数据处理、可视化和进一步标注,适用于音乐符号识别和音乐图谱重建的研究。
MUSCIMA++ is a handwritten musical notation dataset created by the Institute of Formal and Applied Linguistics, Faculty of Mathematics and Physics, Charles University. It contains 140 pages of handwritten music, with a total of 91,255 manually annotated musical symbols and 82,261 explicit relationships between symbol pairs. This dataset supports the training and evaluation of models for symbol classification, symbol localization, and music score assembly, applicable to both individual and joint tasks. The dataset provides open-source tools for data processing, visualization and further annotation, which are suitable for research on musical symbol recognition and music score reconstruction.




