search

rafaelpadilla / review_object_detection_metrics

Object Detection Metrics. 14 object detection metrics: mean Average Precision (mAP), Average Recall (AR), Spatio-Temporal Tube Average Precision (STT-AP). This project supports different bounding box formats as in COCO, PASCAL, Imagenet, etc.
Other
1.08k stars 215 forks source link

How to calculate precision for small, medium and large objects? #70

Closed RicRicci22 closed 2 years ago

RicRicci22 commented 3 years ago

Hi, I am trying to calculate the precision of an algorithm on small medium and large objects. For instance I pose constraints on the area, area<150 ----- small area>500 ----- large 150<area<500 ----- medium

How can I calculate the precision for these type of objects? Because I am confused by the fact that the predictions can have different size, but referring to the same true objects, depending on how good is the detector. In this case, how can I know that a predicted bounding box was meant to detect a small, a medium or a large object?? I am so confused by this...

Thank you!

rafaelpadilla commented 3 years ago

Hi @RicRicci22 ,

In the current version, those values are hardcoded. In a future version these values could be adapted.

In this version the default area thresholds here follows the values defined in COCO, that is, small: area <= 322 medium: 322 <= area <= 962 large: 962 <= area.

You can adapt these values for your needs in the following parts of the code: small medium large

Regards, Rafael

RicRicci22 commented 3 years ago

Hi @rafaelpadilla, thank you for your kind answer. And thank you for the repository! :)

github-actions[bot] commented 2 years ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.