Closed this issue 4 years ago · 1 comments
This is amazing work. Have you tested it on KITTI or CityScapes datasets?
How does it compare to other SLAM pipelines?
Hi @adizhol
Thanks! Currently, gradslam only supports dense RGB-D SLAM pipelines - so we have only tested this on indoor reconstruction datasets.
In our paper, we compare reconstruction (and egomotion) performance against their non-differentiable SLAM counterparts and find we perform on par.