About root pose and camera pose

[s0rrymaker77777]

Hi, thanks for your great work,

I want to ask about the difference between root pose and camera pose. For example, the rigid motion field in lab4d only has root body motion, will you also consider the camera pose separately for this rigid motion?

[gengshan-y]

Hi, we define a field2cam separately for each neural field (background or object),

• for the background, it transforms 3D points from the scene coordinate to the camera coordinate. This is usually referred to as the camera pose (G_bg, e.g. in colmap).
• for object fields, it transforms 3D points from the object coordinate to the camera coordinate. This is referred to as root pose (G_obj) in banmo.
• To get the root pose wrt the background scene, one could use G_bg^{-1}G_o, as X_bg = G_bg^{-1}G_o X_o

[s0rrymaker77777]

Thanks so much for your reply, it helps a lot!