Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Issues with Rebound and Backspin Using Pre-trained Model #9

Open
zhanglbthu opened this issue Jul 10, 2024 · 2 comments
Open

Issues with Rebound and Backspin Using Pre-trained Model #9

zhanglbthu opened this issue Jul 10, 2024 · 2 comments

Comments

@zhanglbthu
Copy link

Hello,
When I use the trained model to reproduce the rebound and backspin action, something goes wrong, like the following:

backspin_gs.mp4
rebound_gs.mp4

Has this happened to any of you, please?
Thank you for your reply.

@zhanglbthu
Copy link
Author

Btw, the trained model was downloaded from the official.

@wyhuai
Copy link
Owner

wyhuai commented Jul 12, 2024

Btw, the trained model was downloaded from the official.

Hi, these issue happens in this code version. There are mainly two reasons: (1) the reference data is not accurate, especially the "rebound" data; (2) the current code uses the first frame to initialize the env. In our follow-up work, using accurate HOI data and applying random frame initialization can significantly improve imitation performance.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants