360 Degree Gait capture: A diverse and multi-modal gait dataset of indoor and outdoor walks acquired using multiple video cameras and sensors

Many of the existing gait datasets are limited by their lack of diversity in terms of the participants (e.g., gender, age, height, weight, ethnicity), recording environments (e.g., recording angles, indoors / outdoors), and availability. Therefore, we present a gait dataset containing 65 diverse participants with both indoor and outdoor environments. The data was acquired using 2 digital cameras and a digital goniometer (used to measure joint angles). Each participant provided 24 walking sequences from a range of viewing angles (360 degrees in 45 degree increments). Each participant also provided an alternative outfit to provide diversity in personal appearance. This dataset will be of value to applications gait identification, human pose estimation, and more.

现有多数步态数据集（gait dataset）均存在局限性：在受试者群体（涵盖性别、年龄、身高、体重、种族）、录制环境（含录制视角、室内/室外场景）以及数据可得性方面缺乏多样性。为此，本研究提出一款涵盖65名多样化受试者、同时覆盖室内与室外录制场景的步态数据集。该数据集的数据采集采用2台数码相机与一台数字测角仪（digital goniometer，用于测量关节角度）。每名受试者完成了24段行走序列，覆盖全360°视角（以45°为间隔依次采集）。此外，每名受试者还提供了另一套着装，以丰富人物外观层面的多样性。本数据集可应用于步态识别（gait identification）、人体姿态估计（human pose estimation）等诸多领域，具有较高的研究与应用价值。

搜集汇总

背景与挑战

背景概述

该数据集是一个多角度、多模态的步态数据集，包含65名多样化参与者在室内和室外环境中的行走数据，使用2台数码相机和数字测角仪采集，覆盖360度视角（以45度递增），并包括不同服装以增强多样性。数据集适用于步态识别、人体姿态估计等机器学习应用，发布于2022年，数据规模为18GB，遵循开放许可协议。