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

What makes the huge performance boost over official implementations? #25

Open
xlliu7 opened this issue Mar 27, 2019 · 1 comment
Open

Comments

@xlliu7
Copy link

xlliu7 commented Mar 27, 2019

Hi, @sunnyxiaohu
Thanks for your excellent and helpful implementation of R-C3D.
I notice that your implementation reports a much higher mAP value on THUMOS14 and wonder what makes the difference.
I guess that you conduct extensive experiments during the implmentation. So according to your experience, what are the main causes?
Thanks!

@coolbay
Copy link

coolbay commented Aug 8, 2019

I have the same question. How did you gain such significant improvement from the original implementation? Is it because you use multiple GPUs to train?

Thank you very much!

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