YGD-mix

A modified version of the Youtube Gesture Dataset. Original data can be downloaded and processed [here](https://github.com/youngwoo-yoon/youtube-gesture-dataset/tree/master) and [here](https://github.com/ai4r/Gesture-Generation-from-Trimodal-Context).

This dataset is used for Audio-visual speaker extraction conditioned on body gestures in [the SEG paper](https://ieeexplore.ieee.org/document/9774925), which the code can be found [here](https://github.com/zexupan/seg).

本数据集为YouTube手势数据集（Youtube Gesture Dataset）的修改版本。原始数据可通过以下两处链接下载并处理：[https://github.com/youngwoo-yoon/youtube-gesture-dataset/tree/master] 与 [https://github.com/ai4r/Gesture-Generation-from-Trimodal-Context]。
该数据集被用于SEG论文（链接：https://ieeexplore.ieee.org/document/9774925）中提出的以身体手势为条件的音视频说话人提取（Audio-visual speaker extraction）任务，相关代码可在[https://github.com/zexupan/seg]获取。