Can I output the pose with an absolute scale?

[942411526]

Dear author， Thanks for your great work, but I have a problem with this, I want to predict the pose from the video. Can I output the pose with an absolute scale?

[guangkaixu]

Hi, thanks for your attention. The output depth maps are scale-invariant, i.e., they are up to an unknown scale, but the scale values are the same for all frames. The output poses are also up to an unknown scale, and the scale is the same as that of depth. But you can try to recover the absolute scale by computing the ratio of optimized depth and metric depth. (Only a few metric depth points are required due to the share of scale values.)