How to get 3D keypoints in order to evaluate in NoW challenge? about deep3dfacerecon_pytorch

[DoubleYanLee]

Hi! thank you for releasing the code for your amazing research! When I evaluate my model (pre_model also be evaluated) in NoW challenge, I also follow the guide from https://github.com/soubhiksanyal/now_evaluation.

python check_predictions.py <predicted_mesh_path>
this code exports the input mesh to predicted_mesh_export.obj . when I open the mesh in MeshLab, I find it not a face shape. it just like that:

I find the page said: "The used mesh loader is unable to load OBJ files with vertex colors appended to the vertices. I.e. if the OBJ contains lines of the following format v vx vy vz cr cg cb\n, export the meshes without vertex colors."

I wonder if the reason is the obj file outputed by test obtained the vertex colors.

Again, thank you for your attention!

[zhih-li]

Hi! thank you for releasing the code for your amazing research! When I evaluate my model (pre_model also be evaluated) in NoW challenge, I also follow the guide from https://github.com/soubhiksanyal/now_evaluation.

python check_predictions.py <predicted_mesh_path> this code exports the input mesh to predicted_mesh_export.obj . when I open the mesh in MeshLab, I find it not a face shape. it just like that:

I find the page said: "The used mesh loader is unable to load OBJ files with vertex colors appended to the vertices. I.e. if the OBJ contains lines of the following format v vx vy vz cr cg cb\n, export the meshes without vertex colors."

I wonder if the reason is the obj file outputed by test obtained the vertex colors.

Again, thank you for your attention!

Hi~ I meet the same issue in the first time with default test.py. On the 66 line in test.py , it save the .obj file using save_mesh()function, and you can edit the function in 215 line in facerecon_model.py, just remove the

vertex_colors=np.clip(255. * recon_color, 0, 255).astype(np.uint8)

In this way, I can get the correct .obj file after executing check_predictions.py.

By the way, did you get 3d coordinates of 7 key points? I'm still a little confused about that.. Thanks for your attention.

[YuDeng]

Hi, we first select 7 vertices on the template mesh (mean shape) as keypoints indicated by NoW, Then we automatically calculate the 3D coordinates of 7 keypoints for each subjects from their reconstructed meshes.

[DoubleYanLee]

Hi! thank you for releasing the code for your amazing research! When I evaluate my model (pre_model also be evaluated) in NoW challenge, I also follow the guide from https://github.com/soubhiksanyal/now_evaluation.
python check_predictions.py <predicted_mesh_path> this code exports the input mesh to predicted_mesh_export.obj . when I open the mesh in MeshLab, I find it not a face shape. it just like that:

I find the page said: "The used mesh loader is unable to load OBJ files with vertex colors appended to the vertices. I.e. if the OBJ contains lines of the following format v vx vy vz cr cg cb\n, export the meshes without vertex colors."
I wonder if the reason is the obj file outputed by test obtained the vertex colors.
Again, thank you for your attention!

Hi~ I meet the same issue in the first time with default test.py. On the 66 line in test.py , it save the .obj file using save_mesh()function, and you can edit the function in 215 line in facerecon_model.py, just remove the

vertex_colors=np.clip(255. * recon_color, 0, 255).astype(np.uint8)

In this way, I can get the correct .obj file after executing check_predictions.py.

By the way, did you get 3d coordinates of 7 key points? I'm still a little confused about that.. Thanks for your attention.

sorry, I just saw your answer. This solved my problem, Thank you!

For the question about getting 3d coordinates of 7 key points, I guess we need to get 3d coordinates of 7 key points in the BFM model(Maybe in the function save_mesh of facerecon_model.py) In 68 landmarks, We could use the landmarks with indices [36,39,42,45,33,48,54] , But I don't know if it works on the problem. Thanks for your attention.

[AyushP123]

Hi @YuDeng,

For computing, the 7 3D locations for the NoW benchmark is the following procedure correct?

1. Use the 7 landmark points from the similarity_Lm3D_all.mat and find the index of the nearest vertices for those in the mean shape.
2. Use the same indices in the final shape to get the 3D landmark points

[wengjincheng]

@DoubleYanLee

I found that just selecting 7 key landmarks by the array that you mentioned, the result of distance calculated by now_evaluate was very large.

[wengjincheng]

Hi @YuDeng,

For computing, the 7 3D locations for the NoW benchmark is the following procedure correct?

1. Use the 7 landmark points from the similarity_Lm3D_all.mat and find the index of the nearest vertices for those in the mean shape.
2. Use the same indices in the final shape to get the 3D landmark points

Hi have you found the correct indices of these 7 landmarks?

[WangXuan2401]

Hi! thank you for releasing the code for your amazing research! When I evaluate my model (pre_model also be evaluated) in NoW challenge, I also follow the guide from https://github.com/soubhiksanyal/now_evaluation.
python check_predictions.py <predicted_mesh_path> this code exports the input mesh to predicted_mesh_export.obj . when I open the mesh in MeshLab, I find it not a face shape. it just like that:

I find the page said: "The used mesh loader is unable to load OBJ files with vertex colors appended to the vertices. I.e. if the OBJ contains lines of the following format v vx vy vz cr cg cb\n, export the meshes without vertex colors."
I wonder if the reason is the obj file outputed by test obtained the vertex colors.
Again, thank you for your attention!

Hi~ I meet the same issue in the first time with default test.py. On the 66 line in test.py , it save the .obj file using save_mesh()function, and you can edit the function in 215 line in facerecon_model.py, just remove the

vertex_colors=np.clip(255. * recon_color, 0, 255).astype(np.uint8)

In this way, I can get the correct .obj file after executing check_predictions.py.

By the way, did you get 3d coordinates of 7 key points? I'm still a little confused about that.. Thanks for your attention.

thanks a lot for your contribution