Hi! Thanks for sharing this great work.
I’m trying to use MANO pose parameters — specifically global_orient with shape (1, 3) and poses with shape (15, 3) — to drive the UHM model. However, UHM appears to use a different skeleton structure from MANO. According to the provided driving_poses files, the parameter formats are different:
# UHM mode:
# ==>> root_pose.shape: torch.Size([1, 3, 3])
# ==>> hand_pose.shape: torch.Size([1, 23, 3, 3])
# ==>> trans.shape: torch.Size([1, 3])
# MANO mode:
# ==>> root_pose.shape: torch.Size([1, 3])
# ==>> hand_pose.shape: torch.Size([1, 45])
# ==>> trans.shape: torch.Size([1, 3])
May I ask how to convert MANO pose parameters from another dataset into the UHM pose representation?
Any guidance would be greatly appreciated. Thank you!
Hi! Thanks for sharing this great work.
I’m trying to use MANO pose parameters — specifically global_orient with shape
(1, 3)and poses with shape(15, 3)— to drive the UHM model. However, UHM appears to use a different skeleton structure from MANO. According to the provideddriving_posesfiles, the parameter formats are different:May I ask how to convert MANO pose parameters from another dataset into the UHM pose representation?
Any guidance would be greatly appreciated. Thank you!