AIST dataset

该数据集包含了13,940个视频，涵盖60首音乐作品、10种舞蹈风格以及35位舞者，旨在用于训练一个由音频驱动的舞蹈生成模型。从该数据集中提取了2,378个片段，并使用不重叠的音乐轨道来划分训练集和测试集。该数据集的规模为13,940个视频，任务是对音乐同步的舞蹈视频生成。This dataset contains 13,940 videos, covering 60 musical works, 10 dance styles and 35 dancers, and is designed for training audio-driven dance generation models. A total of 2,378 clips are extracted from this dataset, with the training and test sets partitioned via non-overlapping music tracks. The total scale of this dataset is 13,940 videos, and its core task is music-synchronized dance video generation.