CtrSVDD

CtrSVDD是由罗切斯特大学等机构创建的大规模、多样化的歌唱声音深伪检测数据集，包含47.64小时的原始歌唱音频和260.34小时的深伪音频，涉及164位歌手和14种深伪技术。数据集来源于公开的歌唱数据集，通过先进的SVS和SVC技术合成深伪音频。CtrSVDD旨在推动歌唱声音深伪检测的研究，解决音乐产业中未经授权的深伪音频问题，保护艺术家和版权所有者的权益。

CtrSVDD is a large-scale and diverse deepfake singing voice detection dataset created by the University of Rochester and other institutions. It contains 47.64 hours of raw singing audio and 260.34 hours of deepfake audio, involving 164 singers and 14 deepfake techniques. The dataset is sourced from public singing voice datasets, and its deepfake audio samples are synthesized using advanced SVS and SVC technologies. CtrSVDD aims to advance research on deepfake singing voice detection, address the problem of unauthorized deepfake audio in the music industry, and protect the rights and interests of artists and copyright holders.