You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi! I was trying the test mode in your code "experiments.py" and I figured out that you don't obtain as output the loss function, in order to see if it is the same as the one in training time.
I have seen in the paper of monodepth that they only obtain disparity, convert into depth and compare with GT.
Why don't you calculate also the loss function?
Thank you,
Luisa
The text was updated successfully, but these errors were encountered:
Hi! I was trying the test mode in your code "experiments.py" and I figured out that you don't obtain as output the loss function, in order to see if it is the same as the one in training time.
I have seen in the paper of monodepth that they only obtain disparity, convert into depth and compare with GT.
Why don't you calculate also the loss function?
Thank you,
Luisa
The text was updated successfully, but these errors were encountered: