Using evaluateFeatureDetector, getRecall, and computeRecallPrecisionCurve in Python

How do I or can I use the following functions evaluateFeatureDetector, getRecall, and computeRecallPrecisionCurve in OpenCV python?

It does not seem to exist in python from what I see.

I appreciate any help!

all true, sad as it is
is it “important enough” to modify the src code & build from scratch ?

Hi Berak,

Thank you for your reply!

In my honest opinion, I do not think it is “important enough” to manually modify the src code and rebuild it. I was looking at the possibility of having some algorithm compute repeatability scores for me! (like with respect to different rotation and scales).

This question arose from my confusion in the definition between matches and correspondences. With reference to the paper “Performance Evaluation of Feature Detectors and Descriptors Beyond the Visible”, there is a paragraph that explains how the repeatability score is calculated, and it stated that (pg 7):

repeatability = number of correspondences / number of features detected in one of the images

Could I find out what correspondences actually mean here? It talks about ellipticals and technical terms that are pretty complex to me.

Does it refer to the number of feature matches after applying matching algorithms such as BFMatcher to two sets of detected features?

I got abit confused as there were two graphs in the paper side by side, and they were “num of matches against different rotations” and “num of correspondence against different rotations” (pg 15).

URL to the paper: (PDF) Performance Evaluation of Feature Detectors and Descriptors Beyond the Visible

imo, "num of correspondence against different rotations” == length of the matches array, and “num of matches against different rotations” is the num of correct matches

related: