sauradip/STALE

Performance under GT proposals

Coder-Liuu opened this issue · 2 comments

image

Thank you for your outstanding work, it's presumptuous to bother you. I would like to ask, shouldn't the performance under mAP@0.5 and mAP@AVG be equal when we use the GT proposal?

Hi,
Although the proposals are GT, the class prediction will be different. So when you combine GT proposals with predicted Class it will be different

Thanks for being able to reply to me.
Although the class predictions are different, the proposal intervals are the same. Therefore, this does not affect the scores at different IoU. When using GT proposals, the scores should be the same for different levels of IoU.

I think if the GT proposal is used, it's just a classification task isn't it?