-
Notifications
You must be signed in to change notification settings - Fork 9.3k
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
Fix gt priority tiny bug #3208
Fix gt priority tiny bug #3208
Conversation
update mmdetection fork
Hi, thanks for the contribution. Have you run the baseline after fixing this? |
Nope. I also want to see the performance gap due to the fix. |
I have tried to re-benchmark the performance with the fix. The AP of Resnet50 does not change (still 37.4), while that of Resnet50 with 0.2-0.5 treated as ignored decreases a bit (from 37.0 to 36.0). Performance of other backbone is still pending. |
An update on the performance.
|
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks for pointing this bug out. Most of configs are quite robust to this modification while the performance with 0.2-0.5 as ignored decreases a bit. I will update this log afterwards.
The original method calculates gt priority has a tiny bug, it actually outputs the rank, not the priority.
E.g. areas=tensor([200., 300., 500., 400., 100.])
The sort_idx would be tensor([2, 3, 1, 0, 4]). But the priority is actually tensor([3, 2, 0, 1, 4]) (Due to the gt areas). So I use an torch.argsort() method to get it.