Runtime of different algorithm steps

[t-fi]

Hi everyone. I just read the glomap paper and wonder what the run time of the different steps in scene reconstruction is. IIUC glomap uses the COLMAP feature extraction and matching output. So any discrepancy in runtime is due to faster geometric optimization?

I would generally be glad if people could share some run times they get on different project scales and hardware :) It probably strongly depends on several factors.

Thanks!