You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi! I'm rencently using your work on some downstream reconstruction tasks. However, I noticed that apart from SH coefficients, other parameters of 3DGS are not likely to be updated? I've checked the requires_grad options and backward.cu in submodules, and it all seems well. Is there any modifications of your code to ensure that the rest parameters of 3DGS are no longer to be unpdated?
Hoping for your reply soon!
The text was updated successfully, but these errors were encountered:
OK, after hours of debugging, I finally know why. There is an option named iter_fix_smpl_init_verts and is set as 1500, indicating that no vertice of human will be changed before 1500.
I just list the answer here in case for others who have the same problem waste time here.
Hi! I'm rencently using your work on some downstream reconstruction tasks. However, I noticed that apart from SH coefficients, other parameters of 3DGS are not likely to be updated? I've checked the
requires_grad
options andbackward.cu
in submodules, and it all seems well. Is there any modifications of your code to ensure that the rest parameters of 3DGS are no longer to be unpdated?Hoping for your reply soon!
The text was updated successfully, but these errors were encountered: