Open JinyongJeong opened 7 years ago
@rdube
If I am not mistaken when reading the descriptor code, and based on Weinmann 2014 ("Hence, it follows that λ1 ≥ λ2 ≥ λ3 ≥ 0 holds for each 3D point X."), it seems that the current code is indeed wrong.
@JinyongJeong thank you for this catch!
@JinyongJeong good catch and thank you for pointing it out! Although this feature is indeed falsely named it seemed to be descriptive and highly selected by the random forest. We will fix this when pushing following demonstrations and features. Thank you!
In this code of eigen descriotor, order of eigen value is e1<e2<e3.
So, I think linearity should be calculated with (e3 - e2) / e3).
(e1 - e2) / e1) also have discriminative feature, but have wrong meaning.
is it right?