Depth-Based In-Bed Human Pose Estimation with Synthetic Dataset Generation and Deep Keypoint Estimation
Abstract
This paper describes a method of estimating the pose of a human in bed only from a single depth image. Such estimation is useful for robotic monitoring of the elderly and the disabled, where their lying posture may indicate illness. While it can address privacy and illumination issues, depth images make the pose estimation problem more challenging. We solve this problem by generating training images with cloth simulation and deep keypoint estimation. We evaluated the effectiveness of the dataset using synthetic and real test images. We also show that adding a small number of real training data improves the results.
Cite
Text
Ochi and Miura. "Depth-Based In-Bed Human Pose Estimation with Synthetic Dataset Generation and Deep Keypoint Estimation." European Conference on Computer Vision Workshops, 2022. doi:10.1007/978-3-031-25075-0_45Markdown
[Ochi and Miura. "Depth-Based In-Bed Human Pose Estimation with Synthetic Dataset Generation and Deep Keypoint Estimation." European Conference on Computer Vision Workshops, 2022.](https://mlanthology.org/eccvw/2022/ochi2022eccvw-depthbased/) doi:10.1007/978-3-031-25075-0_45BibTeX
@inproceedings{ochi2022eccvw-depthbased,
title = {{Depth-Based In-Bed Human Pose Estimation with Synthetic Dataset Generation and Deep Keypoint Estimation}},
author = {Ochi, Shunsuke and Miura, Jun},
booktitle = {European Conference on Computer Vision Workshops},
year = {2022},
pages = {672-685},
doi = {10.1007/978-3-031-25075-0_45},
url = {https://mlanthology.org/eccvw/2022/ochi2022eccvw-depthbased/}
}