BlindWays
收藏资源简介:
BlindWays数据集由波士顿大学、早稻田大学和马里兰大学帕克分校共同创建,是首个针对盲人行人多模态运动基准。该数据集包含11名盲人参与者在真实城市环境中导航的3D运动数据,共计1,029个运动片段和约0.6百万个人体姿态,以及2,058个详细的高低级文本描述。数据集通过穿戴式传感器收集,涵盖了盲人在动态户外环境中的自然运动数据,旨在解决现有3D人体运动数据集缺乏多样性和偏向视力正常人群的问题。BlindWays数据集的应用领域包括自动驾驶和社交机器人等辅助技术,旨在提高这些系统对残疾人运动的预测和响应能力,从而提升安全性。
BlindWays dataset, co-created by Boston University, Waseda University, and University of Maryland, College Park, is the first multi-modal motion benchmark for blind pedestrians. This dataset contains 3D motion data from 11 blind participants navigating in real urban environments, totaling 1,029 motion segments, approximately 0.6 million human body poses, and 2,058 detailed high-level and low-level text descriptions. Collected via wearable sensors, the dataset covers natural motion data of blind people in dynamic outdoor environments, aiming to address the problem that existing 3D human motion datasets lack diversity and are biased towards sighted populations. Application fields of the BlindWays dataset include assistive technologies such as autonomous driving and social robots, with the goal of improving the prediction and response capabilities of these systems towards the movements of people with disabilities, thereby enhancing safety.




