融合和标注的人体姿态数据集

本研究通过融合和标注MOT17和NCLT两个数据集，创建了一个大规模的人体姿态数据集，专门用于捕捉在共享城市区域中人与机器人交互的场景。该数据集包含数千个真实世界的室内外场景，通过YOLOv7网络进行处理，生成了人体姿态的视觉和数值输出，并通过手动标注提供地面实况。数据集的应用领域包括人体姿态预测和社交感知机器人路径规划，旨在解决自主配送机器人在城市环境中安全导航的问题。

This study develops a large-scale human pose dataset by fusing and annotating two datasets, MOT17 and NCLT, which is specifically tailored for capturing human-robot interaction scenarios in shared urban areas. This dataset encompasses thousands of real-world indoor and outdoor scenes. It is processed using the YOLOv7 network to generate visual and numerical outputs of human poses, and provides ground truth via manual annotation. The application fields of this dataset include human pose prediction and socially-aware robot path planning, aiming to address the safe navigation problem of autonomous delivery robots in urban environments.