pycolmap RGBD for pose estimation.

Thank you for the excellent work! I would like to know if it is possible to estimate the pose based on the provided RGBD (RGB + depth) and camera intrinsics. If it is possible, are there any related documents? Thank you very much.

Jul 29 '24 02:07 BJHYZJ

@BJHYZJ Have you found any solution? I am also looking for similar works.

Jul 31 '24 11:07 hiyyg

@hiyyg Hi, I couldn't find it either, it's really hard to find. I'm working on indoor scene reconstruction, and sometimes the accuracy of traditional SLAM isn't sufficient for my needs.

Aug 04 '24 14:08 BJHYZJ

I have the same question.

Oct 25 '24 07:10 guangkaixu

It seems that this issue mentioned a possible solution: https://github.com/colmap/colmap/issues/607#issuecomment-483689208

Oct 25 '24 07:10 guangkaixu

The aforementioned issue is concerned with using external depth inputs for MVS, not for pose estimation. There are multiple ways to use depth in pose estimation: for scoring in RANSAC (not supported by COLMAP) or for 3D-3D alignment (implemented in COLMAP but not exposed to Pyhton yet (PR welcome), but available in Open3D under the name global registration).

Oct 26 '24 09:10 sarlinpe