search

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

About colmap's scale proplem #47

Closed BoMingZhao closed 1 year ago

BoMingZhao commented 1 year ago

Hi, thanks for you amazing work!! I saw that when using the Cambridge dataset, the camera pose and point cloud are obtained from colmap, while the cambridge dataset was originally obtained from visualSFM. I compared the two files and found that the camera pose is not consistent. Whether there will be a scale that will lead to inaccurate evaluation results?

tsattler commented 1 year ago

Can you share statistics of the camera position errors? Do you take into account that Colmap stores translations while VSFM stores camera center positions?

BoMingZhao commented 1 year ago

Can you share statistics of the camera position errors? Do you take into account that Colmap stores translations while VSFM stores camera center positions?

Thank you very much for your reply! I didn't consider it.