Open quetalasj opened 1 year ago
Hi, thanks for your interest.
We haven't tested our method on their SLAM dataset so we don't have a sense which one would be better.
I think the idea of using depth as extra signal is similar, but using which formula of depth loss depends on different tasks.
Hello, I found that iMAP SLAM and Nice-SLAM authors also propose depth losses to increase the speed of the training. Could you kindly explain if your depth loss can be combined with the losses proposed by them? Is there a significant difference between your depth-loss formulation and their or it's just a different notation for the same formula with the same result?