OpenXD-RenFace

RenFace is the only large-scale whole human-head dataset with rich fine-grained hairstyles, as well as phoneme-balanced speech videos produced by the corresponding individual/head. The total number of IDs in this dataset is more than 500. The total number of captured frames reaches 80M, covering 60 viewing angles, with an image resolution of 2k. The dataset contains multi-ethnic, multi-age, multi-ethnic subjects, matched with multiple high fine-grained hairstyles and hair colours, with audio and video collected simultaneously on a reading corpus. The dynamic 4D face dataset can be applied in the fields of 2D/3D digital human face generation, face reconstruction and face capture etc..

RenFace是目前唯一一款大规模全头部数据集，具备丰富的细粒度发型，同时包含由对应个体生成的音素平衡语音视频。该数据集的身份总数超过500个，总捕获帧数达8000万，覆盖60个拍摄视角，图像分辨率为2K。数据集涵盖多族裔、多年龄段的受试者，搭配多种高细粒度的发型与发色，并采用朗读语料库完成音视频的同步采集。这款动态4D人脸数据集可应用于2D/3D数字人脸生成、人脸重建与人脸捕捉等领域。

搜集汇总

背景与挑战

背景概述

OpenXD-RenFace是一个大规模人类头部图像数据集，包含超过500个ID和8000万帧图像，覆盖60个视角和2K分辨率。数据集具有丰富的注释信息，适用于2D/3D数字人脸生成、重建和捕捉等领域。