Giters

danini / graph-cut-ransac

The Graph-Cut RANSAC algorithm proposed in paper: Daniel Barath and Jiri Matas; Graph-Cut RANSAC, Conference on Computer Vision and Pattern Recognition, 2018. It is available at http://openaccess.thecvf.com/content_cvpr_2018/papers/Barath_Graph-Cut_RANSAC_CVPR_2018_paper.pdf

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

About threshold

zzxp opened this issue · comments

commented

Hello ，could you tell me that how to determine these thresholds？
image

commented

Hey! It is a slightly late answer, but:
probability: the confidence required in the results, same as in RANSAC.
fps: e.g. if fps = 10, then the algorithm terminates after 1 / 10 secs.
threshold: is simply the RANSAC threshold
lambda: is the weight of spatial coherence, it is set experimentally
neighborhood_size: it is also set experimentally, the radius of the sphere to determine which points are neighbours.