Usually when using COLMAP the process is to do a sparse alignment, and then run colmap image_undistorter. Does the refinement run before or after this step, or how does it interact with this type of data?
Image undistortion is required only for dense reconstruction, so the refinement does not need this step (we don't run it) and can handle distorted images.
Usually when using COLMAP the process is to do a sparse alignment, and then run
colmap image_undistorter
. Does the refinement run before or after this step, or how does it interact with this type of data?