dockstring

该数据集名为KIT，包含了3,911个运动序列和6,278个文本注释，每个运动序列都关联了1到4个描述。此外，该数据集被下采样至每秒12.5帧，并分为80%的训练集、5%的验证集以及15%的测试集。其规模为3,911个运动序列和6,278个文本注释，任务是对文本进行运动合成。This dataset, named KIT, contains 3,911 motion sequences and 6,278 text annotations, with each motion sequence linked to 1 to 4 descriptive captions. The dataset has been downsampled to 12.5 frames per second, and partitioned into three subsets: 80% for training, 5% for validation, and 15% for testing. The task of this dataset is text-to-motion synthesis.