search

NVIDIA / ContrastiveLosses4VRD

Implementation for the CVPR2019 paper "Graphical Contrastive Losses for Scene Graph Generation"
Other
200 stars 41 forks source link

Issue of AP calculation #31

Closed Scarecrow0 closed 3 years ago

Scarecrow0 commented 3 years ago

Hi, I'm wondering why only take the first rectangle of the precision-recall curve when calculating the AP.

https://github.com/NVIDIA/ContrastiveLosses4VRD/blob/f7539ca1bda75ef7a6bf23713b927b4f19607813/lib/datasets_rel/ap_eval_rel.py#L115

As shown in this line, it seems only take the first different point of the precision-recall curve.

Thanks for your reply.