MMFace4D

MMFace4D是一个大规模的多模态4D（3D序列）人脸数据集，由IEEE TRANSACTIONS ON MULTIMEDIA创建。该数据集包含431个身份，35,904个序列，以及390万帧，旨在解决现有3D人脸动画数据集规模有限或质量不满足需求的问题。MMFace4D具有两个显著特点：一是涵盖了年龄在15至68岁之间的多样化演员和语料库；二是具有同步音频和高分辨率人脸细节的3D网格序列。数据集通过三个同步的RGBD相机捕捉细微的3D面部表情，用于训练高保真、表现力和可泛化的人脸动画模型。MMFace4D的应用领域包括VR/AR、游戏和电影制作，旨在通过提供高质量的数据集推动音频驱动3D人脸动画技术的发展。

MMFace4D is a large-scale multimodal 4D (3D sequence) facial dataset released by IEEE Transactions on Multimedia. This dataset encompasses 431 unique identities, 35,904 sequences, and 3.9 million frames, aimed at addressing the shortcomings of existing 3D facial animation datasets, including limited scale and unsatisfactory quality. MMFace4D boasts two distinct features: first, it covers diverse actors and corpora with ages ranging from 15 to 68 years old; second, it provides synchronized audio and 3D mesh sequences with high-resolution facial details. The dataset captures fine-grained 3D facial expressions using three synchronized RGBD cameras, and is designed to train high-fidelity, expressive and generalizable facial animation models. The application domains of MMFace4D include VR/AR, gaming and film production, with the goal of advancing the development of audio-driven 3D facial animation technology through the provision of high-quality datasets.

搜集汇总

背景与挑战

背景概述

MMFace4D是一个大规模多模态4D人脸数据集，包含431个身份、35,904个序列和390万帧，具有多样化演员和同步音频的高分辨率3D网格序列，旨在推动音频驱动3D人脸动画技术在VR/AR、游戏和电影制作等领域的应用。