Thanks for the great work.
I was trying to evaluate the PiFUHD network on some other synthetic dataset. I encountered issues regarding the affine transformation (majorly scaling factor) of predicted and ground truth meshes which evaluating the metrics (chamfer and P2S distance).
How could you decide/predict the transformation matrix for any dataset?
Also, I looked into your evaluator.py script where you have set some scaling and offset factors defined for both the datasets. But while evaluating the Chamfer and P2S distance you haven't taken any transformation factor into account, and just compared two meshes.
Kindly let us know how can we decide the transformation matrix for evaluating a custom dataset, and what kind of mesh registration technique can we use to do the same.
Thanks for the great work. I was trying to evaluate the PiFUHD network on some other synthetic dataset. I encountered issues regarding the affine transformation (majorly scaling factor) of predicted and ground truth meshes which evaluating the metrics (chamfer and P2S distance). How could you decide/predict the transformation matrix for any dataset?
Also, I looked into your evaluator.py script where you have set some scaling and offset factors defined for both the datasets. But while evaluating the Chamfer and P2S distance you haven't taken any transformation factor into account, and just compared two meshes.
Kindly let us know how can we decide the transformation matrix for evaluating a custom dataset, and what kind of mesh registration technique can we use to do the same.