hi,I am doing some work on point cloud acceleration nerf. But I find the pose generated by colmap is different from the pose provided by the dataset. So observing the point cloud under the pose provided in the dataset, the image projection seen is not consistent with the dataset. How did you solve this problem?
Hi, each run of COLMAP would give different 3D points and poses. We also pre-process the poses in our dataset. It wouldn't work if the point cloud is viewed under the poses given by a different COLMAP run.
hi,I am doing some work on point cloud acceleration nerf. But I find the pose generated by colmap is different from the pose provided by the dataset. So observing the point cloud under the pose provided in the dataset, the image projection seen is not consistent with the dataset. How did you solve this problem?