I wonder if anyone could help me.
I’ve exported the FaceMesh of Metahuman_004 as USD, got rid of the skeleton, and oriented the mesh so Y-axis is up (in Maya) and exported as another USD and dragged into Audio2Face.
But when I do the MeshFitting, it never worked, always stuck on 37.5%. Here’re the last few lines of the log:
2021-04-29 05:45:48 [615,233ms] [Info] [carb.python] [PointPicker] Begin stroke
2021-04-29 05:45:48 [615,307ms] [Info] [carb.python] [PointPicker] End stroke
2021-04-29 05:45:53 [620,709ms] [Info] [carb.python] [PointPicker] Stop picking
2021-04-29 05:46:01 [629,194ms] [Info] [omni.hydra] Destroy mesh instance for /World/metahuman_004_head/group1/Section0
2021-04-29 05:46:01 [629,198ms] [Info] [omni.usd] ResyncedPaths: /World/metahuman_004_head/group1/Section0_fitted
2021-04-29 05:46:02 [629,205ms] [Info] [carb.python] Handle changed 1 prims for layer anon:000001E827DFA670:World1.usd.
2021-04-29 05:46:02 [629,205ms] [Info] [carb.python] Load prim spec tree rooted from /World/metahuman_004_head/group1/Section0_fitted
2021-04-29 05:46:02 [629,205ms] [Info] [carb.python] Handle changed prims for layer anon:000001E827DFA670:World1.usd done.
2021-04-29 05:46:04 [631,351ms] [Error] [carb.python] [c:\users\jqshe\appdata\local\ov\pkg\audio2face-2021.2.4\kit\python\lib\asyncio\base_events.py:1619] Task exception was never retrieved
future: <Task finished coro=<_do_fitting() done, defined at c:\users\jqshe\appdata\local\ov\pkg\audio2face-2021.2.4\exts\omni.kit.charTransfer\omni\kit\charTransfer\commands.py:15> exception=RuntimeError(‘Factor is exactly singular’)>
Traceback (most recent call last):
File “c:\users\jqshe\appdata\local\ov\pkg\audio2face-2021.2.4\exts\omni.kit.charTransfer\omni\kit\charTransfer\commands.py”, line 84, in _do_fitting
register_laplacian_func = register_laplacian_factor(L, landmarks)
File “c:\users\jqshe\appdata\local\ov\pkg\audio2face-2021.2.4\exts\omni.kit.charTransfer\omni\kit\charTransfer\fitting\registration.py”, line 66, in register_laplacian_factor
return partial(_register_laplacian_solve, L, landmarks, landmark0_mask, spla.splu(lhs))
File “c:\users\jqshe\appdata\local\ov\pkg\audio2face-2021.2.4\exts\omni.audio2face.player_deps\deps\py\scipy\sparse\linalg\dsolve\linsolve.py”, line 339, in splu
ilu=False, options=_options)
RuntimeError: Factor is exactly singular
Hi @james.shen ! Thank you for your interest in Audio2Face. To investigate your issue we need to reproduce it. Could you please save your stage (via Audio2Face File->Save As) as a USD file (after you’ve added all the corresponding points before hitting “Begin Mesh Fitting”) and share with us.
Hi @james.shen, thanks for trying Audio2Face. Can you please check if your face mesh prim consists of more than 2 separable meshes (that do not share any edge or face)? With a high chance, the error is related to the input face mesh that you use. It will be easier to see what went wrong, if you provide the .usd file.
Hi @alexandre.messier ,
If you set all the meshes combobox back to “Select Mesh”, and reselect them, does that fix the problem?
Maybe there’s certain combination of mesh selection on the mesh that produce stale info for the UI. I’ll try see if I can repro it here, but meanwhile, if you can provide your repro steps, it’ll be much appreciated.
I am encoutering the same error. Has there been a solution yet? I’ve tried deselecting the Driver and Target options and reselected everything, but unfortunately that didn’t do the trick…
@hinsemartin & @alexandre.messier
the Mesh Fitting panel is not expanding because it silently failed to find the Default Prim in your Usd Stage.
I’ll see how we can avoid this error in the future release, but meanwhile, you can set the default prim on the top node of your mesh by Right clicking on it on the Stage view and choose Set as Default Prim.
Once you set this, try reselect the mesh again in the Meshes panel.
Let me know if this works for you :)
