BowenC/time-signature
收藏资源简介:
Time Signature Detection Dataset是一个基于YOLO格式的目标检测数据集,专门用于识别乐谱图像中的音乐时间签名。该数据集从两个来源编译而成:DeepScores V2的一个子集和来自IMSLP(国际乐谱图书馆项目)的公共领域乐谱,包括肖邦、贝多芬、莫扎特、勃拉姆斯、德彪西等作曲家的作品。数据集包含650张图像,分为381张训练图像和269张验证图像,总共有3,125个时间签名实例。涵盖19个时间签名类别,如4/4、3/4、6/4、2/4、常见时间、6/8、8/4、9/4、13/8、10/4、3/8、7/8、9/8、6/2、4/2、切割时间(alla breve)、12/8、5/4和3/2。数据集结构包括图像文件夹(PNG格式)和标签文件夹(YOLO格式的.txt注释),标签格式为<class_id> <x_center> <y_center> <width> <height>(归一化到0-1)。该数据集适用于音乐符号识别和光学音乐识别(OMR)任务,特别是时间签名检测。
A YOLO-format object detection dataset for recognizing musical time signatures in sheet music images. The dataset is compiled from two sources: a subset of DeepScores V2 and public domain scores from IMSLP (International Music Score Library Project), including works by composers such as Chopin, Beethoven, Mozart, Brahms, and Debussy. It contains 650 images split into 381 training images and 269 validation images, with a total of 3,125 time signature instances across 19 classes, such as 4/4, 3/4, 6/4, 2/4, common time, 6/8, 8/4, 9/4, 13/8, 10/4, 3/8, 7/8, 9/8, 6/2, 4/2, cut time (alla breve), 12/8, 5/4, and 3/2. The dataset structure includes image folders (PNG format) and label folders (YOLO-format .txt annotations), with labels in the format: <class_id> <x_center> <y_center> <width> <height> (normalized 0–1). It is designed for music symbol recognition and optical music recognition (OMR) tasks, specifically time signature detection.



