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

How to use ECON output to train SCANimate #81

Open
ireneisme opened this issue Jul 19, 2023 · 2 comments
Open

How to use ECON output to train SCANimate #81

ireneisme opened this issue Jul 19, 2023 · 2 comments

Comments

@ireneisme
Copy link

ireneisme commented Jul 19, 2023

Thanks again for your work! I still have two question about how to modify the output of ECON so that it can be used for training SCANimate. #51

The first question is that How can I transform the body_pose parameter in SMPL-X to pose parameter in SMPL? I am confused with what exactly those 24 joints are, to be more specific, the 23 body joints which is 24 in SMPLX model.

# SMPL-X (~.npy) produced by ECON
betas :  torch.Size([1, 200])
body_pose :  torch.Size([1, 21, 3])
global_orient :  torch.Size([1, 1, 3])
transl :  torch.Size([1, 3])
expression :  torch.Size([1, 50])
jaw_pose :  torch.Size([1, 1, 3])
left_hand_pose :  torch.Size([1, 15, 3])
right_hand_pose :  torch.Size([1, 15, 3])
scale :  torch.Size([1, 1])

# SMPL (~.npz) needed by SCANimate
transl (3,) 
pose (72,) 
v_cano (6890, 3) 
v_posed (6890, 3) 

In addition, I also have some trouble in removing the invisible faces. I have already checked the query_color function and try to find those invisible part. But I could only remove those invisible vertices, what should I do next to further remove those invisible faces?

# I tried to remove some invisible vertices through this way
vert = vert[visibility == 1.0]

Looking forward to any helpful suggestion😄

@ireneisme
Copy link
Author

The first question is that How can I transform the body_pose parameter in SMPL-X to pose parameter in SMPL? I am confused with what exactly those 24 joints are, to be more specific, the 23 body joints which is 24 in SMPLX model.

# SMPL-X (~.npy) produced by ECON
betas :  torch.Size([1, 200])
body_pose :  torch.Size([1, 21, 3])
global_orient :  torch.Size([1, 1, 3])
transl :  torch.Size([1, 3])
expression :  torch.Size([1, 50])
jaw_pose :  torch.Size([1, 1, 3])
left_hand_pose :  torch.Size([1, 15, 3])
right_hand_pose :  torch.Size([1, 15, 3])
scale :  torch.Size([1, 1])

# SMPL (~.npz) needed by SCANimate
transl (3,) 
pose (72,) 
v_cano (6890, 3) 
v_posed (6890, 3) 

For this question, I have found the 23 = the same 21 body_pose as in SMPLX and 2 for hands. Is it correct? Can I just set both hands to (0, 0, 0) or how can I get the pose parameters for them?

@ireneisme
Copy link
Author

ireneisme commented Aug 4, 2023

The first question is that How can I transform the body_pose parameter in SMPL-X to pose parameter in SMPL? I am confused with what exactly those 24 joints are, to be more specific, the 23 body joints which is 24 in SMPLX model.

# SMPL-X (~.npy) produced by ECON
betas :  torch.Size([1, 200])
body_pose :  torch.Size([1, 21, 3])
global_orient :  torch.Size([1, 1, 3])
transl :  torch.Size([1, 3])
expression :  torch.Size([1, 50])
jaw_pose :  torch.Size([1, 1, 3])
left_hand_pose :  torch.Size([1, 15, 3])
right_hand_pose :  torch.Size([1, 15, 3])
scale :  torch.Size([1, 1])

# SMPL (~.npz) needed by SCANimate
transl (3,) 
pose (72,) 
v_cano (6890, 3) 
v_posed (6890, 3) 

For this question, I have found the 23 = the same 21 body_pose as in SMPLX and 2 for hands. Is it correct? Can I just set both hands to (0, 0, 0) or how can I get the pose parameters for them?

For hand pose params, I set them by transfering the smplx model to the smpl model and calculating their value using the generated .pkl

Update(23/8/9): However, the result is very strange. It seems there are some problems in params and I tried to normalize them, but they still works poor...

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

1 participant