Pressure Eye: In-bed Contact Pressure Estimation via Contact-less Imaging

This is the Pressure Eye (PEye) implementation to estimate in-bed human contact pressure with the bed surface. Both IR and RGB top view image can be employed as vision source in the pressure map estimation.

Env Setup

This project is developed under python3.8/Anaconda 3. Please check requirements.txt for required packages.

Dataset deployments.

Apply the SLP dataset from ACLab SLP
Point --dataroot to where all datasets are stored. The vis-pressure dataset will be under it.
Get the pixel value distribution [statistics](pwrs-phy-ssim-D, extract to misc.

pretrained models.

Please refer to our paper for the the specific configurations.

RGB: pwrs-phy, pwrs-phy-ssim-D

LWIR: pwrs-phy, pwrs-phy-ssim-D

Getting Started

To train (with RGB, physical vector 1, pwrs, ssim and D loss):

python train.py --model vis2PM --mod_src RGB --mod_tar PMarray --n_phy 1 --type_whtL pwrs --lambda_sum 1e-6 --lambda_ssim 10 --lambda_D 1

To test with similar configuration:

python test.py --model vis2PM --mod_src RGB --mod_tar PMarray --n_phy 1 --type_whtL pwrs --lambda_sum 1e-6 --lambda_ssim 10 --lambda_D 1

Other sota approaches are also included in pmScripts:

openPose.sh
memNet.sh
pix2pix.sh
cycleGAN.sh

Comparison with state-of-the-art

Citation

If you find this work helpful, please cite the following work. @article{liu2021pressureeye, title={Pressure Eye: In-bed Contact Pressure Estimation via Contact-less Imaging}, author={Liu, Shuangjun and Ostadabbas, Sarah}, journal={http://arxiv.org/abs/2201.11828}, year={2021} }

ostadabbas/PressureEye