How to get the poses and intrinstics of input video

[zhangyujie-1998]

Thank you for releasing the code. I find the released demo of St4RTrack does not directly output the poses and intrinstics of input videos like MonST3R. I want to know how to get these parameters.

[Junyi42]

Hi,

Thanks for the question! As a pointer, https://github.com/HavenFeng/St4RTrack/blob/main/dust3r/camera_solver.py#L155, one can refer to the gt_intrinsics and pred_pose here to get the predicted poses and intrinsics, which are derived from the predicted pointmaps. We will update the code soon to output the intrinsics and cam poses more explicitly.

Thanks