Question about pose

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! 