Seeing Under the Cover: A Physics Guided Learning Approach for In-Bed Pose Estimation

  • 2019-07-03 23:41:23
  • Shuangjun Liu, Sarah Ostadabbas
  • 24

Abstract

Human in-bed pose estimation has huge practical values in medical andhealthcare applications yet still mainly relies on expensive pressure mapping(PM) solutions. In this paper, we introduce our novel physics inspiredvision-based approach that addresses the challenging issues associated with thein-bed pose estimation problem including monitoring a fully covered person incomplete darkness. We reformulated this problem using our proposed Under theCover Imaging via Thermal Diffusion (UCITD) method to capture the highresolution pose information of the body even when it is fully covered by usinga long wavelength IR technique. We proposed a physical hyperparameter conceptthrough which we achieved high quality groundtruth pose labels in differentmodalities. A fully annotated in-bed pose dataset calledSimultaneously-collected multimodal Lying Pose (SLP) is also formed/releasedwith the same order of magnitude as most existing large-scale human posedatasets to support complex models' training and evaluation. A network trainedfrom scratch on it and tested on two diverse settings, one in a living room andthe other in a hospital room showed pose estimation performance of 99.5% and95.7% in PCK0.2 standard, respectively. Moreover, in a multi-factor comparisonwith a state-of-the art in-bed pose monitoring solution based on PM, oursolution showed significant superiority in all practical aspects by being 60times cheaper, 300 times smaller, while having higher pose recognitiongranularity and accuracy.

 

Quick Read (beta)

This feature is not avaialbe for this paper.