同时收集的多模态躺姿数据集 (SLP)
收藏资源简介:
同时收集的多模态躺姿数据集(SLP)是由东北大学电气与计算机工程系增强认知实验室创建的,旨在解决在不利视觉条件下床内人体姿势监测的问题。该数据集包含来自109名参与者的床内姿势图像,使用RGB、长波红外、深度和压力图等多种成像模式进行捕捉。SLP数据集设计与主流人体姿势数据集兼容,使得最先进的2D姿势估计模型能够有效地使用SLP数据进行训练,性能高达95%的PCKh@0.5。此外,通过包括额外的模态,姿势估计性能可以进一步提高。SLP数据集的应用领域包括医疗保健中的自动床内行为监测系统,这些系统可以在家中或当地医疗中心不显眼地使用,不仅有助于减少医院访问和感染传播的风险,还可以为已经超负荷工作的护理人员带来一些工作负担的缓解。
Simultaneously-collected Multimodal Supine Posture Dataset (SLP) was developed by the Augmented Cognition Laboratory, Department of Electrical and Computer Engineering, Northeastern University (China), to address the challenge of in-bed human posture monitoring under adverse visual conditions. This dataset comprises in-bed posture images from 109 participants, captured via multiple imaging modalities including RGB, long-wave infrared, depth, and pressure maps. The SLP dataset is designed to be compatible with mainstream human posture datasets, allowing state-of-the-art 2D pose estimation models to be trained effectively on SLP data and achieve performance up to 95% PCKh@0.5. Moreover, pose estimation performance can be further enhanced by incorporating additional modalities. The application scope of the SLP dataset covers automatic in-bed behavior monitoring systems in healthcare, which can be deployed unobtrusively in residential homes or local medical centers. Such systems not only help reduce the risks of hospital visits and infection transmission but also alleviate partial workloads for overburdened caregivers.




