Closed pfeducode closed 1 year ago
Same issue here.
Maybe we should use
cam_metrics.update(loss_tensor_to_dict(fov_l1.mean(axis=1), prefix='fov/'))
Hi, thanks for the issue!
This error arises because your camera objects are storing focal_length
as a tuple (f_x, f_y)
instead of just f
. This could be because of a Pytorch3D version update, but it's an easy fix. Following @FavorMylikes's suggestion above, please add the following before line 71:
if fov_l1.dim() == 2:
fov_l1= fov_l1.mean(axis=1)
I've also committed this change to the code. Thank you!
thank you, it works well
The last update dimension of t is [12,2]
Could you give me some suggestions