Not working

[BouOus]

Hello,

I think there is a problem in the code, I followed all the steps. But I do not get a good result

Here is a screenshot of what I get !

Thank you

[nabeel3133]

@BouOus Can you let me know how many total heads and corresponding face shape parameters did you learn the regression matrix on?

[BouOus]

i used 700 heads and [100 params for the LYHM model] and [199 param for the BFM model].

Can you give me your Regression Matrix to test in my own generated BFM model ?

Thank you

[nabeel3133]

Hello @BouOus. Can you send me your Input_Face.obj so I can verify whether there is an issue with the input or not.

Thank you!

[BouOus]

Hi

Here is my Input_Face.obj Input_Face.obj.zip

Thank you!

[nabeel3133]

@BouOus There is something wrong with your input. Try checking the attached "Input_Face.obj" on your regression matrix. Input_Face.zip

[BouOus]

It's working thank you,

Can you give me your line code to generate the Obj BFM model (Input_Face.obj)

Thank you

[nabeel3133]

The attached code snippet extracts a face from BFM Model (BFM Eigenspace)[01_MorphableModel.mat]. However, it extracts the same face every time. In order to generate different BFM faces, you will need to change the shape parameters which is a 199x1 dimensional vector for BFM model.

face_model_dict = loadmat('01_MorphableModel.mat') face_U = face_model_dict['shapePC'] face_mean = face_model_dict['shapeMU'] shape_parameter = np.ones((np.shape(face_U)[1],1)) input_face_tensor = face_mean + np.matmul(face_U,shape_parameter)

To learn more about generating different shape parameters provided only a single image, refer to the following repository: -> https://github.com/microsoft/Deep3DFaceReconstruction

[BouOus]

can you give me your script to generate random BFM and save it into obj. I tried with many ways but I still can not have the same heads for both models.

Thank you

[nabeel3133]

@BouOus Please refer to the following commit: https://github.com/nabeel3133/combining3Dmorphablemodels/commit/714c30a8e8de9eba367c9999b50bf5f7f88d6e14 I have also edited the README.md. Make sure to read it again.

[BouOus]

Thank you for your answer I was sure the code did not work. I tried in many ways and it still produces the same heads.

[nabeel3133]

@BouOus The code will produce the same heads if the regression matrix is learned from only a few head and corresponding face shape parameters or if the shape parameters of the randomly generated bfm faces are close to eachother. Also you won't be able to see the difference in the shape of the head if you see it from the front. You need to open different heads in the same project of Meshlab and see them from either right or left side of the mesh by hiding and unhiding different head meshes. By doing that, you will be able to see the difference.