MP-3DHP: Multi-Person 3D Human Pose Dataset

Multi-Person 3D HumanPose Dataset (MP-3DHP) 是一个基于深度传感器的数据集，旨在促进针对现实世界挑战的多人 3D 姿态估计方法的开发。该数据集包括 177k 训练数据和 33k 验证数据，其中 3D 人体姿势和身体片段均可用。该数据集还包括 9k 个干净的背景数据和 4k 个测试数据，包括多人 3D 姿势。

Multi-Person 3D Human Pose Dataset (MP-3DHP) is a depth-sensor-based dataset developed to facilitate the development of multi-person 3D pose estimation methods targeting real-world challenges. This dataset includes 177k training samples and 33k validation samples, where both 3D human poses and body segments are available. It also contains 9k clean background samples and 4k test samples that involve multi-person 3D poses.

搜集汇总

背景与挑战

背景概述

MP-3DHP是一个基于深度传感器的多人3D人体姿态数据集，旨在支持现实世界多人3D姿态估计方法的研究。它包含17.7万训练样本和3.3万验证样本，并提供9千个干净背景数据和4千个测试数据。