search

facebookresearch / ScaDiver

Project for the paper "A Scalable Approach to Control Diverse Behaviors for Physically Simulated Characters"
Other
152 stars 18 forks source link

Reproducing Results from the Paper #7

Open christsa opened 3 years ago

christsa commented 3 years ago

Hi

I have a couple of questions regarding implementation:

  1. Can you elaborate on how you removed the motions containing interactions from CMU in a semi-automatic manner as stated in the paper?

  2. What were the parameters you used to generate the motion graph (e.g., delta, blend-time, etc.)?

  3. How did you find the "offset based on the minimum height for the contact points in each motion"?

Would you be able to provide the data (motion sequences) that you used for training the control policies? Since the generation of the motion sequences from the motion graph is non-deterministic and without knowledge of the hyperparameters, it makes it hard to reproduce the results reported in the paper.

Many thanks for your help.

Jungdam commented 3 years ago

Please read my answer for your previous question.

And for the specific questions you asked here, I'll leave a comment soon!