You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Is it possible to generate images from other views different from the 16 views used during training?
I assume it is not possible to generate more than 16 images, but I tried generating new views by replacing the camera parameters in the pickle file camera-16.pkl. However, I still get the same images when running generate.py.
I suspect this might be because the camera parameters are over-written by the checkpoint
The text was updated successfully, but these errors were encountered:
Hi, the model does not support the generation of images in other views. It is trained to generate 16 views on the elevation of 30 degrees and evenly distributed azimuth. These fixed viewpoints are hardcoded in the attention layers so that generating other views would be low-quality if we change these viewpoints.
Hi
Is it possible to generate images from other views different from the 16 views used during training?
I assume it is not possible to generate more than 16 images, but I tried generating new views by replacing the camera parameters in the pickle file
camera-16.pkl
. However, I still get the same images when runninggenerate.py
.I suspect this might be because the camera parameters are over-written by the checkpoint
The text was updated successfully, but these errors were encountered: