Hi,
I was trying VGGT-Long's artifacts exported to colmap format and the 3dgs training was completely collapsing, However with VGGT-X, atleast its collapsing like VGGT-Long or DA3 colmap export, but its not the same rendering i do get with colmap or reality capture SfM, camera poses. For the same scenes i was getting very good results with 3dgs mcmc (with colmap or realitycapture)
Thank you.