Alleviating the need for Dynamic CVD?

[asafmanor]

Hi, Given that I use my own depth predictions, is there a way to create poses_bounds_cvd table without running and training CVD?
In particular, how are the 17 parameters constructed?

(Assume I have access the COLMAP outputs)

[asafmanor]

Assuming I am creating poses through COLMAP (wrapper from LLFF)
And have my own metric depth maps

[zhengqili]

Hi, If you have metric depth maps ready, you can align your depth maps with COLMAP pose or vice versa before feeding to Dynibar