URMP数据集

URMP数据集是由罗切斯特大学创建的多模态音乐表演数据集，包含44首古典室内乐作品，涵盖从二重奏到五重奏的不同组合。数据集提供了MIDI格式的乐谱、各个声道的音频录音、混合后的音频和视频录音，以及帧级和音符级的标注文件。创建过程中，面临并克服了保持单独录制的乐器声部同步性和表现力的挑战。该数据集旨在支持音乐信息检索（MIR）任务的开发和评估，以及新型的多模态任务，如视觉辅助的多音高分析和复音颤音分析。

The URMP Dataset is a multimodal musical performance dataset created by the University of Rochester. It contains 44 classical chamber music works spanning different ensembles from duos to quintets. The dataset provides MIDI-format sheet music, audio recordings of individual instrument channels, mixed audio and video recordings, as well as frame-level and note-level annotation files. During its development, the challenges of maintaining synchronization and expressive fidelity of individually recorded instrument parts were addressed and overcome. This dataset aims to support the development and evaluation of music information retrieval (MIR) tasks, as well as novel multimodal tasks such as vision-aided multi-pitch analysis and polyphonic vibrato analysis.