DebSDF number of iterations

[Golbstein]

Hi, thanks for sharing your work. In the paper, you mention that the DebSDF model was trained for 7,000 iterations. I’m curious—does one iteration correspond to a full epoch (i.e., a complete pass over all virtual cameras from all panorama images), or is each iteration defined as a single cube-map image being processed?

If it’s the latter, 7,000 iterations seems relatively low, especially considering each iteration involves sampling only 1,024 pixels from a 384×384 cube-map.

Also, it would be great if you could share the training configuration you used for DebSDF. Thanks in advance!