How to reconstruct nerf_synthetic data using colmap?

[Ru1zhi]

Hi, @Xharlie

Thanks for your code and data. I tried to rebuild the nerf_synthetic data using the default settings of colmap dense rebuild pipeline, but I found that the results are poor and the quality of the generated point cloud is far from the one you provided. Could you please share your reconstruction settings or provide an automatic reconstruction script?

For example, on the lego data, it is clear that the key points of the two sides of the tracks were matched together, causing an error in the reconstruction.

[nexuslrf]

I also had such a problem at the beginning, and this is a problem from COLMAP itself. COLMAP may fail to distinguish the two sides of the bulldozer if you use COLMAP to estimate camera pose. You can fix it by setting the camera pose manually for COLMAP, the result should be better (here is the link).

[Quyans]

hey！ i use colmap too。 and i has the tough question that stuck me for days. How to set the camera pose from cameras.txt generated by colmap?

the Rotation Matrix is just parse quaternion to the matrix? and the translation vector is just the Tx Ty Tz? i will be appreciate if you reply to me. or can you leave your email :> .

[Xharlie]

sorry i forget how did I set colmap, but i think i do use the gt pose and it solve the problems.

[Quyans]

已收到邮件

[Ghul-huan]

@Quyans can you share the email ?