Closed Arjeus closed 10 months ago
Hi Arjeus,
yes, for one scene, all the pairs are used.
(please note that some matches are filtered out if they do not satisfy the https://github.com/facebookresearch/PoseDiffusion/blob/57d644416cd31038c5dacc883b9136e2a609f029/pose_diffusion/util/geometry_guided_sampling.py#L170)
Hi,
I would like to clarify how how may image pairs are computed for their respective Sampson Epipolar Error. Are all combinations used? Or just a subset of it? Thanks.