search

autonomousvision / giraffe

This repository contains the code for the CVPR 2021 paper "GIRAFFE: Representing Scenes as Compositional Generative Neural Feature Fields"
https://m-niemeyer.github.io/project-pages/giraffe/index.html
MIT License
1.23k stars 160 forks source link

confusion about camera and world coordinates #18

Closed MrTornado24 closed 3 years ago

MrTornado24 commented 3 years ago

Thanks for your great work! There is some confusion when I run code:

  1. why would we change the second and first dims of translation when changing object depth and its horizontal position, respectively? What is the relationship between camera and world coordinates?
  2. how do you determine the scale and translation ranges of object to scene, e.g. the face experiments on celebahq? Do we need some prior on that?
m-niemeyer commented 3 years ago

Hi @MrTornado24 , thanks for your interest in our project.

  1. I'm not sure what you mean by this? This only means that the third dimension (z-axis) is pointing upwards which is a common choice when defining a coordinate system.
  2. Yes, you need to define the distribution for this. In this work, we just select min and max bounds and apply a uniform distribution over these ranges.