search

facebookresearch / plush_dataset

This is the "plush" dataset associated with the CVPR 2022 paper "Virtual Elastic Objects".
Other
12 stars 3 forks source link

Forces magnitude #4

Closed antonagafonov closed 2 years ago

antonagafonov commented 2 years ago

Hi guys, Kudos on your great work, In the VEO paper, the optimization problem is set as the total object's Energy minimization, where one of the terms in total object's Energy is the Virtual Work of the external forces. As I see in the files, there is force direction tracking using aruco_ids. How do you determine the force magnitude in your setup?

Another question is also regarding points which are connected to wires, as I see in baby_alien verts_fixed is indeed in ears hands, and legs ending areas. For legs I believe the contact indeed is spread to a lot of points, but what about ears and hands, the wires are connected to a small number of points, but there are hundreds of points defined at each wire connection. Can you please explain why there are much more points defined connected to a wire? I suspect is for optimization stability.

Thanks a lot, AA

image

hsiaoyu commented 2 years ago

Hi! I include a new file "plush_data.pdf" that contains an estimation for the relation between the force and the distance that we are using. I would like to point out that since we didn't take gravity into consideration, the learned stiffness(material parameters) is proportional to external force, meaning that only the ratio between the stiffness and the external force is important but not the absolute value. As for the fixed points, unfortunately, we can only estimate those manually. Like you said, fixing too few points may lead to numerical instability during simulation, since the point cloud is down-sampled to 1/10 of points for simulation.

antonagafonov commented 2 years ago

Got You, Thanks