4
$\begingroup$

My image has 4 ground truth bounding boxes and the prediction just has 3 bounding boxes

  • I need to calculate the quality of bounding boxes predicted.

I know how to calculate IoU / IoBB for 2 bounding boxes ( 1 pred and 1 ground )

But whats the algorithm to decide which pair to calculate for?

Would it be correct to calculate it for all pairs of prediction/ground truth exhaustively?

I am a bit confused here

Improve this question
$\endgroup$
2
  • $\begingroup$ Can you answer it if you were able to figure it out? $\endgroup$
    – Syed Saad
    Commented Oct 8, 2019 at 15:41
  • $\begingroup$ I have also come across this problem now but don't know what is the right approach. I was thinking of going through each bounding box in the ground truth file and calculate the area if there is an intersection with a bounding box in the prediction file. Don't know if there is a better approach. @SyedSaad Were you able to figure out a feasible approach? $\endgroup$
    – Abhishek Tyagi
    Commented May 9, 2021 at 20:36

1 Answer 1

Reset to default
1
$\begingroup$

Yes, normally IoU is calculated for all possible pairs within an image. Then the ones with the highest IoU score are selected. You can take a look at this code if you are not sure how to implement that.

Improve this answer
$\endgroup$
1
  • $\begingroup$ I think this approach somewhat mirrors that of Non-Maximum Suppression, but in some interpretations and scenarios it can be said to grossly under-represent the false positive nature of weakly and non-contributing detections. $\endgroup$
    – matanox
    Commented Dec 2, 2023 at 2:01

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.