POP909

POP909数据集由上海纽约大学计算机科学系音乐X实验室创建，包含909首流行歌曲的钢琴编曲版本。数据集主要包含每首歌曲的人声旋律、主乐器旋律及钢琴伴奏的MIDI格式文件，并与原始音频文件精确对齐。此外，数据集还提供了由专业音乐家手工标注的节奏曲线和通过音乐信息检索算法提取的节拍、调性和和弦注释。POP909数据集旨在支持自动化音乐编曲研究，特别是针对钢琴伴奏生成和基于音频的重新编曲任务。

The POP909 Dataset was developed by the Music X Lab, Department of Computer Science, New York University Shanghai. It contains piano arrangement versions of 909 pop songs. The dataset primarily provides MIDI-format files for the vocal melody, main instrument melody, and piano accompaniment of each song, which are precisely aligned with the original audio files. Additionally, the dataset offers tempo curves manually annotated by professional musicians, as well as beat, key, and chord annotations extracted using music information retrieval (MIR) algorithms. The POP909 Dataset is intended to support automated music arrangement research, especially tasks related to piano accompaniment generation and audio-based music re-arrangement.