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

Cannot reproduce the results on DomainNet when K = 1 #5

Open
Openning07 opened this issue Dec 19, 2019 · 2 comments
Open

Cannot reproduce the results on DomainNet when K = 1 #5

Openning07 opened this issue Dec 19, 2019 · 2 comments

Comments

@Openning07
Copy link

Thank for sharing your project~
Based on the data split and settings of hyper-parameters in the paper, I can obtain similar classification performance on DomainNet when K = 3, but not when K = 1 (e.g., in P to R, I get 73.65 accuracy, instead of 76.1 claimed in the paper; in R to S, I get 59.655 accuracy, rather than 61.0 claimed in the paper).

Do I miss some important details ?

@ksaito-ut
Copy link
Collaborator

Hi, are you using pytorch 0.4.0 to reproduce the results?

@Openning07
Copy link
Author

Hi, are you using pytorch 0.4.0 to reproduce the results?

Yes. And I splited the data based on the txt files in SSDA_MME/data/txt/multi you provided.
That is some kind of weird that in the experiments of one-shot setting, I can obtain similar results (i.e., the difference is smaller than 0.5%) in some adaptation scenarios (e.g., C to S, S to P), while in other cases (like, P to C, R to S), the differences are larger than 1.0%.

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