3DPeople

3dperson是一个大型合成数据集，包含约200万张图像，其中40名男性和40名女性执行70个动作。从4个摄像机视图中捕获每个人的动作序列。此外，数据集对面部、3D骨架、深度、一般图像和光线变化进行分段和注释。3DPerson is a large-scale synthetic dataset consisting of approximately 2 million images. It includes 40 male and 40 female subjects performing 70 distinct actions, with their motion sequences captured from four camera viewpoints. Furthermore, the dataset is segmented and annotated across multiple modalities such as facial regions, 3D skeletons, depth maps, standard RGB images, and lighting variations.

搜集汇总

背景与挑战

背景概述

3DPeople是一个大型合成数据集，包含约200万张图像，涵盖80名人员执行70个动作，从4个摄像机视图捕获，并提供了面部、3D骨架、深度等多方面的注释。该数据集由哈佛大学和Institut de Robòtica i Informàtica Industrial于2019年发布。