GD-ML/MACE-Dance

MACE-Dance是一个用于音乐驱动舞蹈视频生成的大规模数据集，包含约70K个舞蹈视频片段，每个片段5-10秒，总计116小时，涵盖20多种舞蹈类型。数据集由两个互补的子集组成：一个以运动为中心的子集，源自FineDance，专注于专业舞蹈动作质量；另一个以外观为中心的子集，收集自高参与度的互联网舞蹈视频，专注于视觉外观多样性和真实感。数据集旨在支持研究生成既在运动学上合理、视觉上连贯，又与音乐良好对齐的舞蹈视频。

MACE-Dance is a large-scale dataset for music-driven dance video generation, containing approximately 70K dance video clips, each 5–10 seconds long, totaling 116 hours, and covering 20+ dance genres. The dataset consists of two complementary subsets: a motion-centric subset derived from FineDance, focused on professional dance motion quality, and an appearance-centric subset collected from high-engagement internet dance videos, focused on visual appearance diversity and realism. The dataset is designed to support research on generating dance videos that are kinematically plausible, visually coherent, and well aligned with music.