PKU-EPIC / HOTrack

[AAAI 2023] Tracking and Reconstructing Hand Object Interactions from Point Cloud Sequences in the Wild
53 stars 5 forks source link

How to generate SimGrasp Dataset? #8

Closed CAN-Lee closed 3 months ago

CAN-Lee commented 3 months ago

Hi @JYChen18 ,

Thanks for your amazing works. I learn a lot from it.

I would infer that the SimGrasp Dataset is made by first using GraspIt to generate grasp poses and then utilizing those grasp poses to generate corresponding hands by MANO. Is it right?

So, would you mind providing any details in implementation of generation of SimGrasp Dataset? Like what developer tools or environments?

Really appreciate it !

Can

JYChen18 commented 3 months ago

Hi Can,

Thanks for your interest in our work.

At that time, to generate SimGrasp Dataset, we modify the code from mano_grasp, which can directly give you the MANO grasp pose.

However, I recommend you try DexGraspNet from our lab. Compared to GraspIt!, DexGraspNet is faster, more GPU-friendly, and most importantly, easier to use because it is in pure Python.

Jiayi

CAN-Lee commented 3 months ago

Hi @JYChen18 ,

Thanks for your reply. So kind of you.

I think I know how to do and I will try DexGraspNet.

Can