search

tjiiv-cprg / EPro-PnP-v2

[TPAMI 2024] EPro-PnP: Generalized End-to-End Probabilistic Perspective-n-Points for Monocular Object Pose Estimation
https://arxiv.org/abs/2303.12787
MIT License
129 stars 8 forks source link

Theoretical question about Importance-sampling (duplicate from previous repo) #3

Closed harsanyidani closed 9 months ago

harsanyidani commented 9 months ago

Hello! I just realised there is now a new repository so I thought I copy my question from there. (I don't know if the old one is still maintained or not, I hope this isn't a problem.) Issue in previous repository

Lakonik commented 9 months ago

Thanks for the question. I very much agree with the answer from Cross Validated (many thanks to Xi'an), that you don't need to decompose the integrand and this is pretty much the standard approach of Monte Carlo integration (as long as you can find the proposal q, if not, just use a plain uniform distribution).

harsanyidani commented 9 months ago

Thanks, I understand it now!