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

Difference from the APs of FCOS and ATSS from mmdetection #8

Closed
youngwanLEE opened this issue Jul 8, 2020 · 2 comments
Closed

Difference from the APs of FCOS and ATSS from mmdetection #8

youngwanLEE opened this issue Jul 8, 2020 · 2 comments

Comments

@youngwanLEE
Copy link

youngwanLEE commented Jul 8, 2020

Hi,

I'm very impressed by your excellent work.

When I read and compared your experiments about FCOS and ATSS, I was confused by those values you mentioned in Table 2 (a).

FCOS ATSS
mmdet 38.6 39.4
yours(Dirac delta) 38.5 39.2

Could you explain what I missed?

Thanks

@implus
Copy link
Owner

implus commented Jul 8, 2020

Hi, a difference around ~0.1 AP is mainly the randomness fact (e.g., different initialization seeds). For example, the performance of GFL_r50_1x is usually 40.2, but sometimes you can get 40.3.

@implus implus closed this as completed Jul 8, 2020
@implus
Copy link
Owner

implus commented Jul 8, 2020

And for an apple-to-apple fair comparison in ablation study, we run all experiments instead of just copying the results from other papers or repos.

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