3D TalKing Eyes Dataset (TKED)
收藏资源简介:
3D TalKing Eyes Dataset (TKED) 是由厦门大学、中国移动(杭州)信息技术有限公司和香港城市大学联合构建的音频-视线数据集,旨在解决语音驱动的3D眼动动画生成问题。该数据集包含约14小时的高质量音频-网格序列,涵盖了眼动、头部运动和面部运动。数据来源于VoxCeleb2和HDTF视频数据集,通过3D面部重建和3D眼动拟合技术生成。数据集的应用领域主要集中在语音驱动的3D面部动画生成,特别是眼动动画的多样性和自然性生成。
3D Talking Eyes Dataset (TKED) is an audio-gaze dataset jointly constructed by Xiamen University, China Mobile (Hangzhou) Information Technology Co., Ltd., and City University of Hong Kong, which aims to address the problem of speech-driven 3D eye movement animation generation. This dataset contains approximately 14 hours of high-quality audio-mesh sequences covering eye movements, head movements and facial movements. The data is sourced from the VoxCeleb2 and HDTF video datasets, and generated via 3D facial reconstruction and 3D eye movement fitting techniques. The application scenarios of this dataset mainly focus on speech-driven 3D facial animation generation, especially the generation of diverse and natural eye movement animations.




