Does this method can use for Multiview Data Association Like CLEAR?

[zhoupengwei]

No description provided.

[plusk01]

Hi, CLIPPER is a pairwise data association algorithm, while CLEAR and MIXER are for muiltiway association. When using a multiway algorithm based on permutation synchronization (e.g., CLEAR, MatchLift, spectral method, etc), it is necessary to first generate hard (i.e., 0 or 1) pairwise associations, which are then optimized to be made cycle consistent in the multiway setting. That initial pairwise association step traditionally comes from something like the Hungarian method. However, if there is geometric structure in the data, then an algorithm like CLIPPER would be superior to Hungarian because it uses extra information.

On the other hand, the MIXER algorithm does not require initial hard pairwise associations. This saves some initial compute steps because Hungarian/CLIPPER is not necessary.

[zhoupengwei]

@plusk01 Thank you for providing such a detailed explanation, I have found that the source code link in MIXER's paper is currently unavailable Code Do you have any plans to open up the source code of this paper? I would like to compare the differences between your previous work on CLEAR to improve the multi-view feature matching results when encountered the ambiguous visual structures in Structure-from-Motion(SfM) algorithm.