Closed DeepDuke closed 10 months ago
Thanks for the interest in our work.
The poses are refined by registering the scan to ncd_quad_gt_pc.ply
and taking the poses in poses_original.txt
as the initial guess. We use open3d's point2plane ICP with a robust kernel for the registration. The difference between poses.txt
and poses_original.txt
is smaller than 10cm on average.
Thank you!
Hi, what method did you use to refine the poses for the Newer College Dataset? And what's the difference between
poses.txt
andposes_original.txt
? TKS!