---
license: bsd-3-clause
tags:
- speaker extraction
- speech separation
- gesture
- pose
- multimodal
- cocktail party problem
---

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).