zju3dv / Vox-Fusion

Code for "Dense Tracking and Mapping with Voxel-based Neural Implicit Representation", ISMAR 2022

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Using ground truth poses during reconstruction

eriksandstroem opened this issue · comments

Hi,
What is the easiest way to run your code with ground truth poses? I see that there is the config option "use_gt", but that does not seem to influence whether gt poses are retrieved during mapping. I could try myself to hack it, but would like to check with you as well since you know the codebase.

Cheers,

Hi @tfy14esa ,

use_gt is a legacy option that is unused in our release code.
With regard to your question, I would say initializing frame poses from the ground truth ones, then keeping them constant by moving pose optimizers out of the optimization loop is the easiest option.

Thanks, makes sense.