Just out of curiosity

[fb82]

Hi, your paper is very interesting, thanks for sharing the code!

Just out of curiosity. In some of my old experiments (see "noRANSAC for fundamental matrix estimation", Table 5), I found that the geometric epipolar error (i.e. sqrt(e12^2+e21^2)) is usually more robust than the symmetric epipolar error. Did you also try some kind of similar error with MAGSAC?

Thanks again, Fabio