Question about 52 expression param

Question

Question about 52 expression param

abcliguanxi opened this issue 2 years ago · 7 comments

abcliguanxi commented 2 years ago

I wanna get the 52 expression coefficients of each expression for each ID。then we can get mesh_vertices through

mesh_vertices = self.core_tensor.dot(id).dot(exp).reshape((-1, 3))

I would be very grateful if you could give me some advice.

Answer 1 · 2023-07-05T07:37:54.000Z

I would also like to have this information

Answer 2 · 2023-10-21T15:15:55.000Z

Have you found those parameters? I'm also looking for them

Answer 3 · 2023-10-21T16:01:28.000Z

I did not get them but I used their fit3d and passed in all the 3D vertices instead of the 68 landmarks, optimized for the ID and exp for 5 iterations and extracted them. It doesn't reproduce shape and expression accurately but doesn't deviate too much either. Hope it helps! . I wish they would share the exact ID and exp params too since these small errors in each shape would result in larger average error for comparison metric

Answer 4 · 2023-10-21T16:38:23.000Z

Thanks! I'll try to fit with the 3D vertices

Answer 5 · 2023-11-15T14:11:38.000Z

Hi @prathebaselva, which model did you use to fit the 3D vertices? I tried 1.6 but it seems the number of vertices between the model (26278) and the mesh (26317) do not match. I noticed that you also replied in that issue. Did you manage to find a solution?

Answer 6 · 2023-11-15T14:38:51.000Z

I used their vertex 16 to 10 or 10 to 16 npy file to do the conversion. do not remember exactly which one i used.

Answer 7 · 2023-11-15T15:34:36.000Z

turned out it should be 10 to 16. Thank you!