search

Healthcare-Robotics / bodies-at-rest

Code + Data for CVPR 2020 oral paper "Bodies at Rest: 3D Human Pose and Shape Estimation from a Pressure Image using Synthetic Data."
Other
62 stars 8 forks source link

Saving SMPL Ground Truth and Point Cloud File #14

Open DavidTu21 opened 4 months ago

DavidTu21 commented 4 months ago

Dear author,

Thank you for this amazing work! I'm exploring the viz_real_cvpr_release.py script in your repository and find it very useful. I'm interested in saving the real point cloud file with its corresponding SMPL ground truth. Could you please provide some guidance or suggestions on how I can modify the viz_real_cvpr_release.py script?

Your expertise and assistance would be greatly appreciated. Thank you in advance for your time and support.

Best regards, David

henryclever commented 4 months ago

Hi, the ground truth real data for this work does not have corresponding SMPL ground truth. In the paper there is a mesh to point cloud comparison because of this. However, I would encourage you to try and fit SMPL bodies to the point clouds if you are able! It would really help the dataset. We did this in a later work (BodyPressure, PAMI 2021) on a different dataset with point clouds.

DavidTu21 commented 4 months ago

Hi Henry,

I hope this message finds you well. Thank you very much for your prompt reply. Upon further reflection, I realize that I could have familiarized myself more thoroughly with the contents of the paper before reaching out.

I just wanted to clarify a few points regarding the SLP3Dfits repository in your BodyPressure project. If I understand correctly, within the SLP3Dfits repository, there exists a point cloud derived from the SLP depth dataset (along with co-registered other modalities), alongside a corresponding ground-truth SMPL human model. Please correct me if I am mistaken. (I am currently awaiting password access to the SLP original dataset...)

Furthermore, I am interested in obtaining the point cloud file. Would it be appropriate to incorporate a function such as open3d.io.write_point_cloud within somewhere inside SLP-3Dfits/view_fits.py file to facilitate this process? If this question is more suited for that GitHub project, please accept my apologies for any inconvenience caused.

Thank you once again for your assistance and understanding.

Warm regards, David

henryclever commented 1 month ago

Hi, sorry for the very late reply to this !

I believe that the view_fits.py reads the pointclouds/depth data in from the danalab folder in the SLP dataset, rather than a file inside of the SLP3D fits repo. L80 in view_fits.py should commpute the point cloud from depth and camera params and from there you can do whatever you need with it (e.g. write it to disk somewhere.)

Hope things are going OK with your explorations in these repos! -H