Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

How many epochs are required to converge? #4

Open
xianshunw opened this issue Aug 3, 2018 · 2 comments
Open

How many epochs are required to converge? #4

xianshunw opened this issue Aug 3, 2018 · 2 comments

Comments

@xianshunw
Copy link

Hi. Thanks for your contribution.
These days I am training your network. The total loss fluctuates between 1.5 and 3 after 10 epochs while the edge_loss is fixed at 0.00244~. Does this means I have achieved a convergent result? How many epochs are required in your expriment? Or can you share your trained model with us? Your work will work for us very much. Eagerly awaiting your reply..

@Zealoe
Copy link

Zealoe commented Sep 15, 2018

also concerned.

@Huang-Jin
Copy link

Huang-Jin commented Mar 24, 2019

I also faced such problems. I followed the guides from the author, but the result in KITTI eigen split is not very good. I trained the network in KITTI for 20w steps and the performance is terrible, abs_rel is 0.2322.
image
The best performance I achieved is in the 15th epoch with abs_rel 0.1924.
image

May be some details for training and test have not been published.
At least I can't reproduce the performance in the paper now.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants