Cover Song Identification Dataset
收藏arXiv2025-09-30 收录
下载链接：
http://www.music-ir.org/mirex/wiki/
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了用于训练卷积神经网络以识别翻唱歌曲的翻唱歌曲对和非翻唱歌曲对。音频文件以立体声格式制作，采样率为44,100赫兹。该数据集能够生成3,300个翻唱歌曲对和496,200个非翻唱歌曲对。在规模上，数据集包含了330首翻唱歌曲（查询集）、670首对照歌曲、2,113个翻唱歌曲对和2,113个非翻唱歌曲对。任务目标是识别翻唱歌曲。
This dataset contains cover song pairs and non-cover song pairs for training convolutional neural networks (CNNs) to recognize cover songs. The audio files are produced in stereo format with a sampling rate of 44,100 Hz. This dataset can generate 3,300 cover song pairs and 496,200 non-cover song pairs. In terms of scale, the dataset includes 330 cover songs (query set), 670 reference songs, 2,113 cover song pairs, and 2,113 non-cover song pairs. The task objective is to identify cover songs.
