How to reproduce the detector repeatability on the Oxford dataset ?

[djkim3]

Hi,
I tested detector repeatability using pretrained models and the Oxford dataset. But my results was not same with paper's. (My results are half or less than the paper.)

I followed the method of 'Scale & affine invariant interest point detectors' that consider that two points A and B correspond if the error in relative point location is less than 1.5pixel.
(In serction 4.1, you said you followed this paper)(LINK)
A is keypoints of image 1 and B is points of image 2, 3, ...

Could you tell me how experiment the repeatibility test? or Could you upload the code about that experiment?

[Jerome Revaud]

Hi
see issue #8