search

otaheri / GrabNet

GrabNet: A Generative model to generate realistic 3D hands grasping unseen objects (ECCV2020)
https://grab.is.tue.mpg.de
Other
238 stars 30 forks source link

How to get the bps.npz #11

Open anindita127 opened 3 years ago

anindita127 commented 3 years ago

Hi,

Very nice work! I wanted to ask how you created the bps.npz file that we download? Also, you have used the rhand_weight.npy file. Did you get it by pre-processing the GRAB dataset?

otaheri commented 3 years ago

Hi @anindita127, thanks for your interest.

We use this repo to get the bps.npz points. In the README file it explains how to generate random or grid points in a sphere (cube or cylinder as well).

Yes, we get the rhand_weights.npy by getting all contacts between the hand and the objects in the GRAB dataset, and then normalize them to be between 0 and 1. This basically gives you which vertices on the hand are more important for grasping.

I hope this helps.

anindita127 commented 3 years ago

Hi @otaheri thanks for the quick reply. An additional question is how did you create the grabnet_splitname.npz files? They have some extra arguments which I don't find after pre-processing the GRAB dataset (e.g. rhand_data.pt file that is created from https://github.com/otaheri/GRAB/blob/master/grab/grab_preprocessing.py )