Open YandanYang opened 1 month ago
Hi, thanks for your interest. That's a good question.
We didn't have any preprocessing steps as we trained our lightcontrolnet only on random 2D renderings. Thus, there was no need for us to specify a canonical rotation for every object.
Hi,
Thanks for your excellent work! Here is a question about the data processing.
I found objects in Objaverse dataset are not in the same initial pose, especially the rotation angle. I guess that's why your code have the args of
rotaion_y
androtation_x
. How do you deal with the angle problem during training? Since it is impossible to set rotation manually during training.Do you do some preprocess steps to make the rotations uniform?