Hi,

Thanks for your awesome work! I'm trying to reproduce SLAM-Former from scratch but I noticed that your evaluation codes are not released yet.

I believe your evaluation protocol is the same as SLAM3R, but your pre-trained weights work much worse (around 'Acc:10.809, Comp:10.182' for replica room0) than those reported in your paper when I use the same protocols... However, the tracking results look very close (around 0.032m for replica room0).

Can you provide us with your evaluation codes?