Poses pre-computed with ORB-SLAM lead to poor reconstruction

[Miyavistka]

Hello,

I'm trying to run colmap dense reconstruction with camera poses pre-computed with ORB-SLAM2. This tool outputs poes in TUM dataset format, which is the following:

• The format of each line is 'timestamp tx ty tz qx qy qz qw'
• timestamp (float) gives the number of seconds since the Unix epoch.
• tx ty tz (3 floats) give the position of the optical center of the color camera with respect to the world origin as defined by the motion capture system.
• qx qy qz qw (4 floats) give the orientation of the optical center of the color camera in form of a unit quaternion with respect to the world origin as defined by the motion capture system.

Therefore I re-arrange the numbers so that they match images.txt required format. I also provide the intrinsics in cameras.txt and an empty points3D.txt. 

However I found that dense reconstruction quality is much better without the pre-computed poses, whereas with the given poses colmap can reconstruct fewer points. 

Has anyone else tried this orb-slam + colmap pipeline? Could it be that I'm not performing some necessary  transformations? I have also tried applying an inverse transform to ORB_SLAM output, but that didn't change COLMAP output much.

Many thanks,

Elena Kosheleva