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

About STL10 top-1 accuracy #6

Open
GA-17a opened this issue Dec 22, 2020 · 2 comments
Open

About STL10 top-1 accuracy #6

GA-17a opened this issue Dec 22, 2020 · 2 comments

Comments

@GA-17a
Copy link

GA-17a commented Dec 22, 2020

Hi, thanks for your share. But I have a question about your STL10 top-1 accuracy. Is the accuracy '70.1' fine-tuning from your pre-trained model 'resnet-18_40-epochs' or training from scratch for 40 epochs?

@hafnerfe
Copy link

hafnerfe commented Feb 3, 2021

@GA-17a if you look in eval/eval.ipynb, it achieves a testing accuracy of 70.1%.
However, the model .../resnet-18-80-epochs/.../model.pth is loaded.
When I use the exact same script to pretrain the model using 40 epochs (i.e. training from scratch), I only achieve an accuracy of 66.5%.
Seeing the other open issue, other people seem to have the same problem. Maybe there's an error in the documentation.

@GA-17a
Copy link
Author

GA-17a commented Feb 4, 2021

@hafnerfe Thanks for your reply, and it solved my question.

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

2 participants