baegwangbin/MaGNet

Loss of D-net

yxt7979 opened this issue · 2 comments

yxt7979 commented

Hi, Excellent work!
I found the loss of D-Net is nll = (torch.square(mu - gt_depth) / (2 * var)) + (0.5 * torch.log(var)) in here.
I have two questions below:

  1. will the loss turns negative as log(1e-5) < 0 (take 1e-5 for example)
  2. how long does the training step of D-net take for each dataset?

Thank you for your reply!!

baegwangbin commented

Hi, thanks for your interest.

  1. Yes, the loss becomes negative and this is not a problem.
  2. We trained the networks for 5 epochs for ScanNet and 10 epochs for KITTI.
yxt7979 commented

Thanks!