Question about pose

[lyg1597]

Hi, thanks for the fantastic work! I am currently trying to apply the method to my own monocular images and I am wondering what should I provide for poses. Should it be camera to world transformation or world to camera transformation? For the poses I provide, will they also get optimized? Thank you so much!

[jashshah999]

Pose drift in MonoGS is because it lacks global optimization. gtsam-splatfactors runs pose optimization inside iSAM2 with loop closure detection — eliminates the accumulated drift.

https://github.com/jashshah999/gtsam-splatfactors