Forums

5th UG2+ challenge (CVPR 2022) Track 1.1 - Object Detection in the Hazy Condition Forum

Go back to competition Back to thread list Post in this thread

> Should change the dryrun data?

I think it is better if we use another dryrun data without given the gt bounding boxes.
It is weird or just me think that mAP >= 99.0 is a little bit off?

Posted by: longpham3105 @ March 8, 2022, 10:44 a.m.

We release the GT labels for dry-run data for the participants to better debug their model's performance.
You are not supposed to use the dry-run data's label in any model's training.
We noticed that there are some 95+ score but these scores are useless, aren't they?

Posted by: wuzhenyu_sjtu @ March 9, 2022, 5:07 a.m.
Post in this thread