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 CUB-200-2011's accuracy #1

Open
shiyan-cui opened this issue Oct 28, 2022 · 1 comment
Open

About CUB-200-2011's accuracy #1

shiyan-cui opened this issue Oct 28, 2022 · 1 comment

Comments

@shiyan-cui
Copy link

Thanks for your work and sharing your codes! I reproduce your code three times on RTX3090*4 entirely following the instruction,but I just got 91.4% as the best accuracy of CUB-200-2011.Could you analyze the problem about this? Thank you for your help。

@learn-in-practice
Copy link

learn-in-practice commented Oct 30, 2022

Thanks for your interest to our work. As stated in the paper and the code introduction, we conduct all the experiments on NVIDIA GeForce GTX 1080 Ti GPUs with python 3.7.7, torch 1.5.0, torchvision 0.6.0, and apex package (installed without --cpp_ext). Different GPU platforms, python package versions, parallel operations, etc., may cause the fluctuation due to function calculation ways, parallel calculation order, and floating point precision. We rearrange the code to make it clear. Check the above content, retry the code, and perhaps slightly tune some training settings suitable to the code operation environment may help.

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