ANTHROPOS-V
收藏资源简介:
ANTHROPOS-V是由罗马大学研究团队创建的一个合成逼真视频数据集,专为人群体积估计(CVE)任务设计。该数据集通过GTA-V游戏引擎生成,模拟了多种城市环境中的大规模人群,并首次提供了每个场景中个体的体积标注。数据集包含每个个体的SMPL形状参数、关键点以及体积信息,数据量较大且具有高度逼真性。数据集的创建过程涉及对GTA-V默认角色的3D网格进行修改,以使其身高和体重分布与现实世界的人类统计数据一致。该数据集的应用领域包括人群管理、公共安全、基础设施压力评估等,旨在通过RGB图像估计人群的总体积,从而解决人群拥挤和结构过载等问题。
ANTHROPOS-V is a synthetic photorealistic video dataset developed by a research team from Sapienza University of Rome, specifically tailored for the Crowd Volume Estimation (CVE) task. Generated via the GTA-V game engine, the dataset simulates large-scale crowds across diverse urban environments, and is the first to provide volume annotations for individual pedestrians in each scene. It contains SMPL shape parameters, keypoints, and volume information for every individual, boasting a large scale and exceptional photorealism. The dataset creation process involves modifying the 3D meshes of GTA-V's default in-game characters to align their height and weight distributions with real-world human statistical data. Application scenarios of this dataset include crowd management, public safety, infrastructure stress assessment and other fields, with the core objective of estimating the total crowd volume through RGB images to address issues such as crowd congestion and structural overloading.

- 1ANTHROPOS-V: benchmarking the novel task of Crowd Volume Estimation罗马大学 · 2025年



