Issues using Mast3r for relative pose estimation on two images

[bakuljangley]

I've run into some issues where I would really appreciate suggestions on how to use your method:

I ran tests on a couple of sets of images taken from my phones camera. The matches were filtered, I only use top n_matches (hyper-parameter in my code) to select 'n_matches' number of points with the highest confidence scores.

For some reason, the returned translation vector (from cv2.solvepnp()) is in the right direction but smaller than expected. Do you recommend scaling the translation ? If so, how?