HIBER

该数据集名为HIBER，包含了来自同一房间内10位不同角度的人类主体的多视角雷达热图。此外，数据集还分为两个子集：“WALK”，其中包含一个主体的数据，以及“MULTI”，包含多个主体的数据。在规模上，“WALK”子集包含73.5K数据帧，“MULTI”子集包含70.8K雷达帧。该数据集的任务是进行目标检测和分割。

The dataset, named HIBER, comprises multi-view radar heatmaps of human subjects captured from 10 distinct angles within a single room. The dataset is further split into two subsets: the "WALK" subset, which holds data from a single human subject, and the "MULTI" subset, which contains data from multiple human subjects. In terms of scale, the "WALK" subset consists of 73.5K data frames, whereas the "MULTI" subset includes 70.8K radar frames. The core tasks supported by this dataset are object detection and segmentation.

搜集汇总

背景与挑战

背景概述

HIBER是一个多领域毫米波人类行为识别数据集，包含10种环境、10名志愿者、4种动作和3种遮挡情况下的400GB预处理数据，支持人体位置跟踪、行为识别、姿态估计和轮廓生成等任务。数据集通过FMCW雷达和光学相机采集，并提供了详细的RF热图、RGB图像及多种标注信息。