MelodySim

该数据集是一个注重旋律的音乐相似性数据集，它通过对Slakh2100 MIDI数据集进行增强，加入了每首歌曲的变化版本，同时保持了旋律元素的不变。该数据集为每首歌曲包含三种变化，旨在保持旋律一致性的同时，改变其他音乐方面。规模上，该数据集包含了6,272个完整的音频音乐文件，包括原始作品及其各种变体。该数据集的任务是基于旋律相似性进行音乐抄袭检测。

This dataset is a melody-focused musical similarity dataset, which is augmented based on the Slakh2100 MIDI dataset by adding varied versions of each song while retaining the original melodic elements. It provides three variations per song, designed to maintain melodic consistency while modifying other musical dimensions. In terms of scale, the dataset contains 6,272 complete audio music files, encompassing both original works and their various variants. The core task of this dataset is music plagiarism detection based on melodic similarity.