Using sfm model on 7scenes without depth

cvg / pixloc

Back to the Feature: Learning Robust Camera Localization from Pixels to Pose (CVPR 2021)

Apache License 2.0

735 stars 92 forks source link

Using sfm model on 7scenes without depth #38

Closed MisEty closed 2 years ago

MisEty commented 2 years ago

Hi, Thanks for your sharing work. I noticed that the demo on 7scenes need the reference_sfm model "sfm_superpoint+superglue+depth". But I cannot download the rendered 7scenes depth images and only have the sfm model builded from RGB images. Can I use such reference sfm model directly on the pre-trained model?

sarlinpe commented 2 years ago

Yes definitely, but the results will not be as accuracy as with depth. Why can't you download the depth images though? We used the renderings available here. The pipeline is implemented here.

MisEty commented 2 years ago

I can download the rendered 7scenes depth images today. Maybe just some network failures and I will try sfm model with depth again. Thanks!