tjiiv-cprg / EPro-PnP

[CVPR 2022 Oral, Best Student Paper] EPro-PnP: Generalized End-to-End Probabilistic Perspective-n-Points for Monocular Object Pose Estimation

Home Page:https://www.youtube.com/watch?v=TonBodQ6EUU

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

A question about equation(1) to equation(2).

TheCuriousJoe opened this issue · comments

Hi HanSheng, very excellent work!
In equation (2), likelihood fuction p(X|y) is defined as .
We can view p(X|y) as the joint probability of p(X1|y), ... , p(X2|y). So .
||fi(y)|| is reprojective error and its value from 0 to infinity. Let x be ||fi(y)||, . So p(Xi|y) is not a pdf.
Is my induction correct? If my induction is correct, statement p(X|y) is not proper here. And following equation can't use Bayes theorem to get p(y|X).
Perhaps i'm splitting hairs, but it really confused me.

Yes the likelihood is not a PDF. Likelihood has nothing to do with density. Only after applying the Bayesian rule can the posterior become a PDF.

Yes the likelihood is not a PDF. Likelihood has nothing to do with density. Only after applying the Bayesian rule can the posterior become a PDF.

The likelihood function is not as normal as usual. Can u give me more detail about how to get equation(2) from equation(1)? Or there is not a strictly math induction?

If you strictly consider the likelihood to be the conditional probability density of X, then Eq. (2) is incomplete because a normalization factor is missing. But this is OK because it's being used as a likelihood and its absolute scale doesn't matter.

Thanks a lot. Excellent work again and wish I could figure out all the equations and intuitions.